Every window, every control.

01What TapeLeader is

TapeLeader prepares multitrack sessions before mixing, for any DAW. Give it the folder, or the archive, that a client or a band sent you, and it gives you back the tracks ready to start work:

• Checked: it finds empty tracks, copies, fake stereo, channels in opposite polarity, clipping, mains hum, bleed and noise, fake bit depth and mixed sample rates.
• Identified: it knows what each track is (kick, snare, lead vocal, backing vocals, guitar…) among 46 types, even when the names come abbreviated, in another language or misspelt.
• Fixed: it joins L/R pairs in stereo, turns what isn't stereo into mono, aligns phase between mics of one source, cleans useless lows with the right filter for each track (zero phase where mics must keep summing, minimum phase on hits where pre-echo would be heard), removes mains hum by subtracting it, and puts a gate where it's needed.
• Levelled: it brings each track to a working level by its type (0 VU, by peaks, or a starting balance by style), under a peak ceiling it never crosses.
• Ordered and named: it renames with your format, numbers by family and writes WAVs with metadata DAWs understand.
• Ready to open: it writes a Reaper project and a .dawproject (Bitwig, Studio One, Cubase 14+) with folders by family and colours, the key, the markers and the tempo changes from the client's notes, plus a report of everything it did.
• Explained: every decision says why it was taken, with the data measured on that track (the INFO tab of the track, under the table).

The originals are never modified. Everything is written to a new output folder. And the project can be non-destructive: the gate as editable automation, the FX Chains open live in the DAW without printing, and a copy of the originals at hand (chapter 15).

It works completely without AI. AI is optional: if you set it up, it decides better where the rules hesitate (odd names, the client's notes, your template…) and never touches what is measured in the audio (chapter 11).

By default TapeLeader applies no "colour" process (tape, saturation, compression): that's the engineer's call, and the engineer's plugins. Its own processes only clean and prepare. If you want, each type of track can go through an FX Chain of your VST3 plugins.

How it decides on each track (why that cut, why that gate), with examples by instrument and references, is in How it decides.

02Installing and starting

Requirements: Windows 10 or 11, 64-bit. Python is installed by uv on its own. (A proper installer is on the way; for the beta, these steps.)

1. Install uv, once

   In PowerShell: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Run test_on_windows.bat

   In the TapeLeader folder. It installs everything, runs the automatic tests, scans your plugins and checks the local AI if you have one.

3. Run Create shortcut.bat, once

   It puts TapeLeader, with its icon, in the Windows Start menu, on the desktop and in its own folder; from the Start menu you can pin it with a right click. Remove shortcut.bat takes it away (the app and its settings stay). If you move the TapeLeader folder, run it again.

4. Open it

   From the Start menu, or with a double click on TapeLeader.bat.

Optional:

• 7-Zip: needed to open RAR and archives split into several files. ZIP, TAR and 7z work with nothing else.
• Ollama: to use a free AI running on your own computer (chapter 11).

When it starts, TapeLeader reopens the last library and the last session you were working on, with everything you'd decided.

[SCREENSHOT: the main window, just opened]

03The basics

The loading bar

At the very bottom of every window that prepares audio (the main one, Align, FX Chains, Strip Silence, Clipper, FixEQ, Track Presets, miniMix and Rough mix; in Track Presets, in the bottom row next to Apply), a thin line says when something is being prepared and what: "Preparing AFTER…", "Loading… (2 of 3 plugins)", "Rendering…". When it's known how much is left, it fills; if not, a stretch runs from side to side. Every control stays usable meanwhile: it doesn't block anything. Anything that takes less than a tenth and a half of a second isn't shown.

04The main window

The window always opens maximised (there's a lot of data in sight); you can resize or minimise it like any other. With two monitors, it opens on the one where you left it. The other windows (Track Presets, FX Chains, Strip Silence, Align, miniMix…) open where and at the size you last gave them.

[SCREENSHOT: the main window with an analysed session, its zones numbered]

Top bar

Each button carries its icon with its name underneath, in the order you work: which session, measure it, decide what to do with it and, at the far right, render.

If the window is narrow, what doesn't fit in the bar goes into »; the four listening buttons always stay in sight.

The header of every window is the same (Align, FX Chains, Strip Silence, Clipper, miniMix and Rough mix): on top, in boxes, what you're looking at (the track with its type, the group or the session, and the figure that matters), the view tools if there are any (Wave | dB, ◀ Hit / Hit ▶) and the player (▶, LOOP, BEFORE, AFTER…; in Align, also PROCESSED / UNPROCESSED). ▶ goes last, on the right, with the time inside it. Below, the track or the whole song with the box of the stretch; the stretch's times go next to the box.

The session picker

Sessions don't take a column on the left: the track table has the full width of the window. To change session:

• Click the session's name in the bar: the list of the library's sessions opens. On top there's a search box: type part of the name and press Enter to open the first that matches, or pick one with the mouse.
• Or use ‹ and › for the previous or the next.

In the list, each session says whether it's a folder or an archive (and whether it's extracted), how many tracks it has, the tempo if known, and whether it brings a reference mix. If two archives are the same session it says "Copy of…". In front goes a dot: hollow if you've never opened it, filled once you have (what you decided is saved), and green once you've rendered it.

What you can do with a session

Right click on any session of the list, or ··· next to the search box for the open one:

• Alternatives: several ways of preparing the same session (other processes, other faders, other FX Chains), each with its name. The ticked one is what you see, it saves itself and it's the one that renders. New alternative… copies the current one under another name. With more than one, the name goes next to the session's ("Lights On · Soft gate") and each renders into its own folder.
• Start over: the current alternative goes back to what the program proposes. What you had is saved first as another alternative: nothing is lost.
• Remove from TapeLeader…: deletes everything TapeLeader did with the session (what you decided, its alternatives, the measurements, what DeBleed learnt and, if it came zipped, the unzipped copy, usually around 1 GB). Back to the hollow dot. Your library and your renders aren't touched, and what you'd decided is set aside just in case.
• Hide from the list: it doesn't show in the list, in the search or with ‹ ›. Nothing is deleted. Show hidden (N), at the foot of the list, shows them in italics; right click on one → Show in the list brings it back for good.

TapeLeader never deletes anything from your library: if a session is surplus, remove it yourself from its folder, or hide it.

The session header

Above the table, in one row:

• Tempo (♩ 120 bpm): the one in the notes or the Readme; if they don't say, the one measured in the tracks (chapter 6). Click to pick between the notes', the measured one, half, double, or type another. If the beat isn't steady it shows "≈".
• Key, in letters ("Ebm"; "E" alone is major): measured in the audio, from the notes, or set by you. If the music isn't clearly major or minor (blues, rock on fifths) it says so instead of guessing.
• The session summary: how many tracks, their format (bits and kHz) and their length.
• The preset in use. On the right, the box "Ask for a change and press Enter…": write what you want ("take the gate off the toms", "the room in mono") and press Enter. The AI turns it into actions of the app and shows them for you to confirm (chapter 11).
• Notes…: what was read from the Readme and other documents of the session (TXT, Markdown, RTF, Word or PDF), with a box to paste the client's email or message. "Read again" interprets it without AI; "Interpret with AI" understands any language and any way of writing. From there come the tempo, the metre, the key, the good take, what not to use, the markers ("the solo starts at 2:10") and the tempo changes, which go into the Reaper project and the DAWproject.

Below, the summary (with a coloured border): the fixes already in the plan, said with names ("Vox copy.wav is not used, audio identical to Vox.wav"; empty tracks, fake stereo, pairs joined, mics aligned…) and the warnings that need attention, like clipping. It can be closed for good with the × on its right; it comes back with Settings ▾ → Summary above the table.

The track table

One row per track and one column for each thing that will be done to it, to see the whole session at a glance. Left to right: who the track is, what was measured that matters, what it gets at render and what it will be called. Column titles, like every label in the app, are in English in every language.

Hide inactive (N), above the table, appears when there are tracks that don't go to the render: it takes them out of view. Nothing is lost: the same button shows them again.

How to read the process columns. What can be put on and taken off is a button with its value inside ("70 Hz", "−6 dB", "Ø", "−1.12 ms"):

• Lit button: applied at render.
• Filled button: lit, but with a different value from the one TapeLeader proposes (you retouched it). The mouse says so.
• Grey button: measured and proposed, or you took it off; not applied. One click lights it.
• A dash (—): doesn't apply to that track.

The titles with ▸ are the columns with their own window: double-click the title to open it on the selected track. A single click on a process title takes you to the PROCESS tab.

Controls. Knobs, faders and sliders work the same on every screen: drag (a knob, up or down), Ctrl while dragging to go ten times slower, the wheel for one step and double-click to go back to the default value (or, if the control has a proposal from the program, to that). Numbers with arrows can be dragged too. To compare, the same three buttons everywhere: BEFORE, AFTER and GAIN MATCH. AFTER has its own colour, so you see at a glance that what plays is processed.

Columns your way. Right click on the table header (or Settings ▾ → Table columns) to choose which are shown. Hidden by default, in case you want them: Format, Duration, Sample PEAK, Act. RMS, LUFS, Active, Gaps and Plan (and Ø and Hum). Default columns resets the view.

The selected track: INFO, PROCESS and ASK

Under the table is always the track you have selected: on the left, a box with three tabs, INFO, PROCESS and ASK; in the middle, the spectrum and, below it, the waveform of the whole track; on the right, the IN / OUT meter and the Monitor out. Each thing has its place, the same on every track, and there are no scroll bars.

INFO is what you need to know about the track, read-only. Selecting a track always comes back here.

• On top, the icon of its type, the file name, its type and version and its format ("MONO · 44.1 kHz · 24 bit · 3:20").
• WARNINGS: the same as the Warnings column, each in its tag and with its three colours. The data and where to look, on hover.
• PROCESSING: the tags of every process, always the same and in the order of the table. Lit, the track carries it, with its value ("Gain +2.0 dB", "HPF 50 Hz"). They inform; to change something, PROCESS.
• ANALYSIS: what was measured on the track, in small boxes: Pitch (a drum's fundamental) or Lowest note; Range and Mostly (from which note to which a voice sings, and where it mostly is); Hits and Decay −20 dB; Sibilance; Top end; Level, PEAK, Hit peak and, on a drum mic, Bleed: what leaks into it most.

PROCESS holds the controls of that track, in eight cards, always in the same place and almost in the order the track plays: on top ALIGN, DEBLEED, HPF, GATE and STRIP; below FixEQ, CLIPPER and FX Chain (STRIP plays after FixEQ, but sits on top because it fits there and leaves room below). In the miniMix, when they go in two rows, it's the order of the signal: Align, DeBleed, HPF and Gate; FixEQ, Strip, Clipper and FX Chain. They are the same cards as in the miniMix. What the program proposed shows when you hover the button of each card's title, and the small ↺ in its top right corner brings the whole card back to the proposal. What you touch here is that track's, above what the program proposes and what Track Presets says for its type, and it's saved with the session.

Next to the FX Chain card, Emphasis and Macros: the Gain of each band and the four knobs and two switches of its FX Chain, with that track's values (chapter 12; Macros are in progress).

ASK is for asking about that track. An AI answers, only with the plan and measurements of the track; it doesn't listen to the audio. Without an AI set up, the field says "AI is off".

The reason for everything shows on hover (why that cut, why that gain, why it's linked to other mics). It isn't written in the box: whoever wants it has it one gesture away.

Notes and keys are written in letters throughout the app (C, C#, D, Eb…; "Em" is E minor, "F#3" is F sharp in the third octave), as in any DAW or tuner.

On the right:

• The full spectrum, 20 Hz to 20 kHz: the average of where the track plays (thick line) and the peaks (thin). Tilted 4.5 dB per octave, as in Pro-Q or SPAN, so a balanced sound looks roughly flat. Marked on it: the fundamental or the lowest note, the resonances of the warnings, where the S's fall and the plan's filters, with what they take shaded. The noise floor shows dotted only when it matters.
• The waveform of the whole track, below, in two tones as in a DAW. On top, a mark at its moment for each S that's too loud, each plosive and each clipping point.

Play the track, A/B. ▶ plays the selected track; while it plays, it's ■ to stop. BEFORE (A) and AFTER (B) choose what you hear, switching without a cut. B is the track as it will come out, with everything in the plan, the same steps as the render. A is the original file. GAIN MATCH (on by default) raises or lowers A to sound as loud as B, so you hear the process and not just "louder"; while it's working it out, the button pulses. The space bar does the same.

• It starts where there's audio: just before the track's first note, not from the silence at the start.
• A click on the waveform puts you at that moment. When it stops, the cursor goes back to where you clicked, as in a DAW.
• If you pick another track while it plays, the new one plays from the same place: useful for comparing mics of one take without stopping.

The meter and Monitor out. Two bars: IN, the track as it comes, and OUT, with its processes. As in a DAW: the light bar is the peak, the solid bar inside is the RMS and the line the highest peak of the last second and a half. Below, Monitor out: the volume you hear at, −24 to +24 dB. It's for listening only: it doesn't change the render.

The table's boxes rule what you hear. If while it plays you untick the gate or the HPF, or change something in Track Presets, B is rebuilt and swapped without stopping.

Session warnings

Almost everything TapeLeader finds belongs to one track and is on its row. What belongs to the whole session (the mains at 50 or 60 Hz, two copies of a track, L/R pairs, mics treated as one source, what the client's notes say, the preset the AI proposes, what TapeLeader can learn from your corrections) is under the ⚠ N Warnings button, top right, next to the notes.

It opens a window with a table of what you need to look at, ordered by instrument: track, type, the warning and, if there's something to apply, what its box does (ACTION). A ticked box is applied at render. Without a box, a mark says where it comes from: ✎ the notes, ✦ the AI, ⚠ or ● what was measured. Double-click a row to go to its track.

What the program fixes on its own isn't listed here (HPF, gate, Strip Silence, Align, hum, DC, stereo pairs…): it's in its column and in INFO.

05Step by step

1. Open the library

   With Library ▾ → Open library…

2. Pick a session

   Click its name in the top bar. If it wasn't analysed, it analyses itself. If it's an archive, press Analyze: it's extracted to the cache (the archive isn't touched) and analysed.

3. Check the types

   In the Type column. Fix what's wrong with a double-click: TapeLeader offers to remember the word that explains it, to get it right next time. With an AI set up, the session is reviewed when it opens.

4. Look at the notes, the tempo and the key

   At the top. If the client sent you an email, paste it in Notes….

5. Go through the table's columns

   At a glance, what gain, HPF, gate, Strip Silence, Clipper, phase and FX Chain each track carries. Tick or untick what you want, in its box or in the Warnings window, which also has what doesn't belong to one track. INFO has the data and, on hover, the reason.

6. Adjust Track Presets

   If the preset doesn't fit: levels, filters, gate and FX Chains, listening to a track before and after while you move the controls.

7. Retouch single tracks

   In PROCESS, or with a right click.

8. Open the miniMix and hear it together

   Decide what can only be heard together (the kit's polarity against the overheads, whether a gate shows, how much of each) and leave the starting faders and pan.

9. Render

   Choose folder and format. When it's done you can open the Rough mix to hear what was rendered, retouch faders and pan, and take them to the Reaper project or the DAWproject.

10. If you want, texts for the client

    Settings ▾ → AI → Texts for the client… gives a summary of what arrived and an email asking for what's missing.

Everything you decide saves itself: close and come back, and the session is as you left it.

06The analysis: what's measured and what it warns about

Each track is read once, in blocks (however long it is), and measured:

• Sample peak and true peak (4× oversampling).
• Active RMS: only where the track plays, so silences don't fool it.
• Integrated LUFS and momentary maximum (ITU BS.1770), and loudness range (EBU Tech 3342).
• Activity (percentage with signal) and level of the gaps (bleed or noise).
• DC offset per channel.
• Clipping: flat runs of samples at the maximum, and where the first is.
• Mains hum: frequency (50 or 60 Hz), level and in which part of the quiet stretches it shows.
• Relationship between channels in stereo tracks (identical, nearly identical, in opposite polarity).
• Real bits (a 24-bit file that is really 16, say).
• Octave bands, spectral centroid, attack rate and crest factor (they help to know what each track is).
• A fine low-end spectrum (about 3 Hz resolution, up to 1 kHz), separately where the track plays and in its gaps, to place the HPF.
• Notes: the lowest note each track really plays, the notes at each moment (chromagram, with tuning correction) and the bass note. They serve the HPF and the key.
• A fingerprint of the audio to find exact copies.

Each track's sheet

Besides that, TapeLeader measures on each track what an engineer looks at before touching a filter or a gate. It shows in the INFO tab and it's what the warnings, the automatic gate and the choice of version by the audio use.

The type of voice isn't stated: it says "compatible with". Bass, baritone, tenor, contralto, mezzo and soprano overlap by more than an octave, and singing, a man can't safely be told from a woman. Nor does it need to: the treatment doesn't depend on the voice type but on each track (the HPF on the lowest note it sings; a de-esser on where that voice has its S's).

Warnings by track type

Warnings don't process anything: they say what there is and where. Two levels: review (in the Warnings column) and information (in the plan only).

Vocals (lead, double and backing):

The rest, by resonances: a band that sticks out 6 dB or more from what's around it (review at 9 dB), with the name of the zone it falls in:

Mixing guides don't agree on frequencies (a kick's "box" goes from 150 to 600 Hz depending on who you ask), so the zones are where a measured peak is looked for, not fixed figures to touch.

Worth knowing about these warnings: the thresholds come from two sessions and from what makers and magazines publish, and nobody has checked them by ear yet. "Low thumps" on backing vocals recorded as a group can be feet or the kick leaking in: the moment is marked for you to listen. Breaths, mouth clicks and headphone click aren't detected yet.

Warnings and their fixes

The list also proposes: an HPF per track with its cut and its reason (9.2); the type's gate where the type asks for one (8); phase and polarity alignment between mics of one source (9.1); a gate by the measured bleed on drum mics (below). With AI: suspicious tracks (one that is the whole mix, one with only bleed), which version to use and doubtful phase (11). Always unticked.

L/R pairs: a stereo pair, or two doubled takes?

"Gtr L" and "Gtr R" can be two mics on one amp (a pair: joined into one stereo track) or two different takes of the same part, panned apart (left as two mono tracks). The names don't tell them apart; the audio does. TapeLeader compares the two tracks in many stretches and measures:

• the coherence of the waveform (0 to 1): high if it's the same performance caught by two mics, nearly zero if they're two takes;
• the delay between them and how much it varies from one stretch to another: two mics on one source have a fixed delay; two takes, one that wanders by several milliseconds.

A stereo pair is joined by default; two doubled takes are proposed unticked, and each is processed on its own. If the measurement falls in no-man's-land, the AI, if you use it, decides with the names and the figures.

Different level between L and R. If one side sounds lower, "Match L/R" is proposed, unticked: it may be a preamp with another gain or the real stereo image (the hi-hat nearer the left overhead). If the difference is nearly constant through the song it says so; if it changes a lot from moment to moment, it warns that it's the image and better left alone.

Mics of one source with no common name. With the same measurement, TapeLeader finds mics of one source even if they're called "Fet47" and "Ribbon": same waveform and fixed delay. Then they get the same gain and filter, and their phase is checked.

Tempo

The tempo is measured in the tracks themselves, without reading the audio again (it uses the envelope the analysis keeps, one value every 10 ms):

1. The tracks that keep the beat

   The click if there is one; if not, drums and percussion; if not, all.

2. The period that repeats most

   Their attacks, over many beats in a row: so the tempo comes out to the hundredth.

3. Does it hold?

   Checked half a minute and a minute later. If it holds, the session was cut to a click and the tempo is exact (within 0.12 bpm of a round value, it's that value). If not, the tempo moves: an approximate value with "≈", or none if there's no clear beat.

Half or double can't be told apart by measuring (68.5 or 137?): it picks the one easiest to count, near 115 bpm, and the tempo button offers half and double. Checked on real sessions whose Readme states the tempo: 104, 115 and 137 bpm, all three exact.

If the notes give a tempo that isn't the measured one, nor its half or double, it warns you. With a measured, steady tempo, the render can also write a click track.

Key

TapeLeader measures the overall key of the song with the tracks of instruments that play notes (not drums or effects): it recognises the chords at each moment (major, minor and fifths) and the bass note, tries each key for how well those chords fit its functions (tonic, dominant, subdominant, and the borrowings typical of rock like ♭VII or ♭VI), and the one that best explains the whole song wins, with a confidence. If the major and minor third sound almost equally (blues, rock on fifths), it doesn't force a choice: it says, for instance, "D, neither clearly major nor minor".

Bleed in the drums

In a kit each close mic also picks up the others. TapeLeader measures how much of each source gets into each mic, in dB against the mic's own hits: −6 dB is nearly as loud as its own; −25 dB you barely hear. With that it decides where the gate goes and with what threshold. Each mic's INFO shows what leaks in most ("bleed: tom 1 at −17 dB").

1. Find the hits

   On each mic. Hits that fall on several at once are one hit.

2. Give each hit its source

   The one whose mic is closest to its peak.

3. Listen where the mic's source is silent

   For one mic and another source, the hits of that source where the mic's own doesn't play: what sounds in the mic then is the bleed.

4. Only what's reliable

   Real bleed follows the strength of the hit, so it barely changes from hit to hit. If it varies a lot or there are fewer than 8 hits, it says "?". If two sources nearly always play together (hi-hat with snare), it marks "≈".

Overheads and room aren't measured: they pick up the whole kit on purpose.

The gate goes only where it's needed. It's off on every type by default, but with the bleed measured TapeLeader ticks it on its own on the close mics of kick, snare and toms when what leaks in is audible and sits with margin under their hits:

With the bleed measured, the gate's threshold isn't the automatic one (meant for noise): it's set between the bleed and the hits, a little nearer the bleed, because cutting a soft hit is worse than letting a little bleed through.

SC: the other drum's sidechain (kick and snare). The Gate card of a kick or a snare has a second button, SC, and a knob, SC Range. It's another gate, before the normal one, that listens to the other drum's mic (the snare in the kick, the kick in the snare): when the other plays and only its bleed is in the mic, it brings it down to the level of what's between hits, never more than SC Range (set on each track between 4 and 12 dB by how much leaks in; with ghost notes, 8 at most), and comes back as the hit dies away. Your hits go through, the soft ones that stand out too.

With Gate off, SC is off too. Mics of one source (in and out, top and bottom) close together. It doesn't take the hi-hat out of the snare, and when kick and snare play together it doesn't close.

DeBleed: taking out what leaks in, inside the hit too. It has its card, DEBLEED, between ALIGN and HPF, and its column between HPF and GATE. It only applies to the close mics of kick (In, Out), snare (Top, Bottom) and each tom. Its button turns it on (with what it had, or −12 dB if it never had anything) and off; inside, two knobs:

• Amount, from Off (top, by default) to −30 dB: the most any frequency of what leaks in from the other drums comes down. With −12 dB nothing comes down more than 12 dB. The first value to try.
• Attack, 0 to 30 ms (10 by default): in the first milliseconds of each hit of the mic, the highs DeBleed had taken are given back, so the attack doesn't lose brightness.

Unlike the SC and the gate, which open or close the whole mic between hits, DeBleed separates by frequency and also inside the hit: the kick and the hi-hat in the snare while the snare plays. It uses no AI: it learns from the session itself how each drum sounds in the main mics of the kit. It sounds after Align and before the HPF, the SC and the gate, and works the same with the gate on or off.

• Only on mics with bleed. If what leaks in is more than 30 dB under the mic's own hits, it says No bleed next to the button and it can't be turned on.
• Calculated once per session. Learning how each drum sounds takes a few minutes (a 4-minute song with six close mics, about 4), so it doesn't happen on its own. Until then the button says Not calculated. The first time you turn DeBleed on, a window says which mics get cleaned, how long it will take on your computer, and a tip: it's calculated with the current Align, so make big Align changes first. What it learns is kept.
• If you change Align afterwards. Inverting a polarity, or moving a mic up to 1 ms, doesn't need a recalculation. If a mic moves more, DeBleed keeps sounding with what it learnt and its button says Other Align: click it to recalculate. It never recalculates on its own. At render, if a track carries DeBleed and the session isn't calculated, it's calculated then.
• If the session has few loose hits of a drum, that track goes in careful mode (it removes less) or isn't touched; the card says so. It needs a snare top mic and a kick mic in the session.

Measured on a real session, in an early test: with the snare at −16 dB in the tom 1 mic, the gate was open 1.4 % of the time (the tom's hits) and closed with the snare playing.

The analysis cache

The analysis is kept in the cache folder. When a new version of TapeLeader measures something more, the old analysis no longer applies and the session is analysed again when you open it.

07What each track is

The 46 types (plus Unclassified)

How it decides

1. By the name

   It understands names in Spanish, English, Italian, German, French and Portuguese (Cassa, Rullante, Gesang, Bumbo, Chimbal, Sala, Raum…), abbreviations (Kik, SN, OH, BV, Gtr…), variants (In, Out, Top, Bottom, DI, Amp), doubles (DT, Dbl) and typos. Mic models count too: D112 or Beta91 point to a kick, R121 to a guitar amp.

2. By the audio

   If the name says nothing, or contradicts what plays, the spectrum, the attacks and the level help decide or mark it as doubtful.

3. By the other tracks

   The session's context helps: mics of one source are recognised as a group and the two sides of an L/R pair get the same type. The AI, if you use it, sees all the names at once: with a "Kick In", it reads "Sub" as a kick.

4. With AI, optional

   Only for the doubtful ones (chapter 11).

It also groups the mics of one source and makes a clean name for each track (ROOM LEFT - M81 → Room Left, KICK IN SM91 → Kick In). And each track has a role: the good one, an alternate take, a printed effect or a guide. Only the good ones count as safe mix material; leaving out the others is proposed unticked.

Change the type, and let it learn

Double-click the Type column and pick another. What you change by hand wins over everything and is saved with the session; everything that depends on the type is recalculated: level, HPF, gate, FX Chain and name.

TapeLeader also offers to remember the word that explains the correction ("Gonzo = kick"). Tick it and next time it gets it right on its own, also without AI. What it learnt is seen and deleted in Learned rules….

In the same drop-down, under each type that has them, are its versions ("Kick · In", "Tom · Floor"…): to give a track the version its name doesn't say, or take it away.

08Track Presets

The most important screen in TapeLeader: almost everything that's done at render is decided here. They are your settings for any project, for every session, not just the open one. It's a normal window, and everything is handled with knobs and buttons (double-click a figure to type it). Where there are a few options to choose from (Measure, Mode, dB/oct, Phase, Expander, the character of Clipper) there's one button that says the current one: a click goes to the next, the wheel goes back and forth, and a right click opens the list.

On the left, the preset and the list of types; on the right, the selection, in cards.

[SCREENSHOT: Track Presets, with the kick selected]

The track's path

Above the cards, in one line, the order in which things are done to a track of the selected type, with what each step is worth for that type. Steps that don't apply are dimmed (DeBleed only shows on kick, snare and toms) and the two levels are highlighted:

1. Align
2. DeBleed
3. HPF
4. Hum
5. Gate
6. FixEQ
7. Strip
8. Input level
9. Ultrasonic LPF
10. Clipper
11. FX Chain
12. Final level
13. WAV

Without an FX Chain, "Input level" is just "Level" and "Final level" is dimmed: it's the same.

The preset

Top left, next to TRACK PRESET, you pick the preset: a complete set of settings for the 46 types. The list goes in three groups: Technical preparation (the usual before mixing), Starting mix (optional) and Yours. Next to it: Save… (what you see, with your changes, as a preset of yours), Delete (yours only) and Undo (back to the preset as it was). Under it, the selected preset is explained, with links to the sources it's based on.

The list of types

On the left are all the types, grouped by family, with these columns: Tracks (how many of that type in the open session), Level, HPF, Gate, Strip and FX Chain. Types not in the session show in grey, but they're edited just the same. What differs from the preset is highlighted. Only this session's types hides the rest.

The order of this list is the order of your tracks. With ↑ and ↓ you move what you've selected: a type, several, or a whole family. That's how tracks come out in the main table of any session, how they're numbered in the output if you use {n}, how they sit in the miniMix and how they go to the DAW project. The original file names never change. Default order goes back to the factory one: drums, bass, lead vocal, backing vocals, acoustic and electric guitars, keys, strings, winds and effects.

Several types at once (Ctrl or Shift + click) or a whole family (click its name): changing a setting changes only that setting on all of them.

Versions of a type

A kick recorded inside and outside, a snare with a mic on top and below, or a bass by DI and by amp are the same type of track and ask for different things. That's what versions are for: each is a complete treatment (level, filters, gate and FX Chain) hanging from its type in the list.

Which version each track carries. The file name says: "KickIn", "Kick Inside" or "Kick Batter" are In; "Kick Out" or "Kick Front", Out; "SnareUp" and "Snare Top", Top; "SnareDown", "Snare Btm" or "SN Under", Bottom; "Rack Tom", "Tom Hi" or "Tom Mid", Rack; "Floor Tom", "FT" or "Tom Low", Floor; "Bass DI", DI; "Bass Amp", "Cab" or "Mic", Amp. Where the name doesn't say and the audio does, the audio says: a tom is Rack or Floor by its fundamental (Floor under 100 Hz). In and Out, Top and Bottom, or DI and Amp aren't decided by the audio: there, the name rules.

Vocals have no versions. Their treatment comes from each track. If you want to treat a group of voices differently, create the version yourself ("With another name…").

They appear on their own. The versions in the session show in the list under their type, with their type's settings. As soon as you touch something in a version, that change is its own. If you leave it equal to the type again, it follows the type again.

The gain between versions. Mics of one source get the same gain, and that still holds between In and Out as long as they have the same level target. Give them different targets (In at −16, Out at −22) and each is levelled to its own. With the HPF, the same.

The cards on the right

On top, the icon and name of the selection, its type preset and Undo the selection. Below, the path and the cards: Level, DeBleed, HPF, Gate, Strip Silence, FX Chain, FixEQ and Clipper, and Listen. Each card knows what the preset says: hovering its title shows "Preset: …", the ↺ in its corner brings it back, and a green dot on each knob marks that value.

Level

• Measure: RMS, LUFS, PEAK or Off.
• Target: the level the track is written at.
• Ceiling: the true peak the track will never pass, even if it doesn't reach its target.

With an FX Chain there are two levels (9.6): the track goes into the chain at the level the chain asks for and, at its output, it's measured again and left at this final level. The input level belongs to the chain and is changed in FX Chains.

HPF

Option: A (the lowest of the track loses 1 dB at most; the default) or B (the cut right there); each track can change it on its HPF card. dB/oct: Auto, 12, 18, 24 or 48. Phase: Auto (by track), Zero or Min. There is no LPF for preparation: it was removed from the whole app. The ultrasonic LPF is something else and always goes (9.7).

Gate

The gate fits each track. With AUTO on, on kick, snare, toms, hi-hat, percussion and claps: hold and release come from how long that track's hit takes to die (hold, half of what it takes to fall 20 dB; release, the whole of it), kept between three quarters and three times the type's values; the detection filter centres on the drum's measured fundamental. At least four of its own hits are needed to measure; if not, the type's values apply.

On drums the values are cautious on purpose: a gentle expander that brings things down 4 dB at most on kick and snare, 6 on toms, hi-hat and percussion. It reduces the bleed without leaving holes and without taking a slightly softer hit. On all: hysteresis 6 dB and lookahead 2 ms (3 ms on lead, double and backing vocals). Synth bass uses the Rest values.

Strip Silence

Right after the gate, a small card like the FX Chain one (9.9): its button (on, Auto: proposed on tracks of the type with long silences; off, not offered), the type's Strip Silence preset (four come with the app: Vocals, Drum hits, Percussion, Instruments, and yours under them), what that preset carries, and ···, which opens the Strip Silence window with the type's values on one of its tracks.

FX Chain

The FX Chain of plugins applied to that type (chapter 12). Its plugins show on hover. Below, Emphasis (the Gain of each band for this type) and the Macros for this type; the green dot and double-click go back to the chain's. ··· opens the FX Chains editor.

Listen

Pick a track of the selected type and press ▶ (or the space bar): the original (BEFORE) and the result with the settings you see (AFTER) play in a loop, even before you press Apply. Every time you move a control, AFTER is recalculated and swapped without cutting the sound. GAIN MATCH brings BEFORE to AFTER's loudness. LOOP repeats the stretch. The stretch can be chosen: drag the box over the small waveform, stretch it by its edges (3 to 40 seconds) or click outside it. It starts on the track's loudest 12 seconds.

FixEQ

The FixEQ button (on = proposed on tracks of the type), Tone, Low, High and FixReso (at the bottom, Auto: what the program proposes on each track; at the top, Off) and ···. What it does and how it decides, in 9.11.

Clipper

The Clipper button (on = proposed on tracks of the type), Mode, the character (Soft, Medium, Hard, Tube), Shave (at the bottom, Auto: the dB each track asks for), PEAK MATCH, a drop-down with the Clipper presets and ···. In 9.10.

Type presets

The type preset button saves and recalls the settings of a single type of track (its whole treatment: level, DeBleed, HPF and its option, gate, Strip Silence, FixEQ, Clipper, FX Chain with its Emphasis and Macros), apart from the general presets. Saved as "Kick metal", "Vocal with soft gate"…; from any type they can be used on any other.

For every track

At the bottom, in the same row as Apply: Same gain for the mics of the same source (on by default) and Ultrasonic LPF: always. When you press Apply, the session's proposals are recalculated. Types whose settings changed get their boxes back as the new settings say; on the rest, what you'd ticked or unticked by hand is respected.

09The processes, in detail

At render each track goes through these steps, in this order:

1. Common format
2. Hum
3. Align
4. DeBleed
5. HPF
6. DC
7. Gate
8. FixEQ and FixReso
9. Strip Silence
10. Gain
11. Ultrasonic LPF
12. Clipper
13. FX Chain
14. Final level
15. Edge fades and WAV

The hum is removed first, measured over the whole track, before Align and the filters. DeBleed, when a track carries it, sounds after Align and before the HPF. The DC offset is subtracted as the file is read, before Align; only the optional very low DC filter runs at its place in the list. Every track comes out the same length and lined up in time (by the BWF time stamps of the originals), so they start at the same point when imported into the DAW.

9.1 Phase and polarity alignment

In each group of mics of one source (kick in/out/sub, snare top/bottom, bass DI/amp, several mics on one amp) one is the reference and each of the others is measured for what it needs to sum with it. It's measured on the hits, three ways. The first two compete and the one that makes the sum gain more wins; the third comes in when neither gives an acceptable fix:

• Polarity first, then time. What an engineer would do: flip it (Ø) and see if the sum gains body. The sum is compared hit by hit with the mic as it is and inverted. If inverted gains at least 1 dB and 7 hits out of 10 say so, it's inverted; then, with the polarity set, a fine time adjustment (under 1 ms) is looked for, used only if it adds something. The case of a snare top and bottom.
• Delay and polarity together, by correlating the attacks (GCC-PHAT), accurate to a fraction of a sample. Right when the mics are at different distances: bass DI and amp, kick in and out.
• By the sum. What you do with the milliseconds knob: slide the mic (up to 2 ms) and stop where the hole in the mids fills without the lows going. The program counts each octave by what you hear, not by its energy, so 3 or 4 dB more at 500 Hz counts even if the lows don't change. It's proposed if the sum gains like that on 7 hits out of 10 and no band that sounds loses more than 1 dB. On a kick three zones can't lose more than half a dB: its fundamental, the knock around 0.7–1.5 kHz and the click (about 4–10 kHz). With three mics on the kick, each moves to sum with the other two together.

Then it looks far, up to 6 ms (two metres between mics). A big move only replaces a near one if it clearly gains more (0.3 dB, once the cost of moving that much is discounted) and passes the same vetoes. When the best move falls where the correlation of the attacks says (within 0.3 ms and with the same polarity), two different measurements are saying the same thing: that's the real delay between the mics. Then it's proposed even if an octave that doesn't rule loses something (up to 2 dB), because the attacks land together.

It only proposes what improves the sum. The proposal is dropped if there are no attacks to measure, if they're already aligned, if the measurement is unreliable (under 50 %) and moving by the sum doesn't gain enough either, if the sum doesn't improve, or if it gets worse by more than 1 dB in a band that sounds. An octave 20 dB under the loudest one counts neither way.

The Align window

Double-click the Align column title (with a track selected), ··· on the Align card, or the Align button. The space bar plays and stops.

On top, three boxes: the group's name (groups are picked in the list on the left) and the player: PROCESSED / UNPROCESSED, ▶, LOOP, BEFORE (the sum unaligned), AFTER (aligned) and GAIN MATCH. Below, the reference mic whole, with the box of the stretch: drag outside it to mark a new one (as short as a quarter of a second, to adjust up close), move it, or stretch it by its edges (up to 40 s). A click outside doesn't change it: playback starts from there. The graphs, the waveforms and the listening are those of that stretch. By default, the reference's fullest 8 s.

The stretch is for seeing and hearing; the options are measured on the whole track. What Align corrects is a fixed delay and polarity for the whole file (the distance between two mics doesn't change mid-song), so the best measurement uses every hit.

For each mic that isn't the reference:

• Apply (ON): whether it's corrected or left as it is.
• Options: the buttons USER, AUTO, A, B…
• Invert (Ø): the polarity. The first thing to try.
• Move: a knob (±5 ms; ±10 ms in Drum kit) and a box with the figure (up to ±20 ms, steps of 0.01). Negative is earlier; positive, later.
• Monitor: the listening level of that signal in this window.
• Below, in one line, what the option in place is and what the sum gains with it.

The options. Each mic has several possible sums and you switch between them with their button, playing too, as between BEFORE and AFTER:

• USER is yours: what the controls say (Ø and milliseconds). The only one you edit.
• AUTO is the recommended one, with its explanation.
• A, B, C, D are other sums that also gain, each a different way: the best move without inverting, the best inverting, polarity only and, on kick and bass, the recommended one with the phase rotated (φ). Each line says what it gains and, if it loses in some band, where and how much: "the sum gains 3.2 dB at 125 Hz, but loses 0.9 dB at 250 Hz · total +1.5 dB".

Three gestures: click another to put it in place; click the one in place again to go back to the one before (AUTO, yours, AUTO, yours, without moving the mouse); touch a control with AUTO, A… in place and it's copied into USER to retouch.

What you leave in place when you press Apply goes to the render, and in the table the Align button says which option it carries: "AUTO −3.36 ms", "USER −3.44 ms", "A Ø +2.46 ms".

While you move the knob or tick Ø, the sum by bands (BEFORE and AFTER), the waveforms of two hits (BEFORE on top, AFTER below) and, if it's playing, AFTER change together, without a cut. Under the bars, the summary by zones: Lows (40–100 Hz), Low mids (125–315 Hz: the body, and the mud too), Mids (400 Hz–2 kHz), Highs and Total. It doesn't say whether it sounds better: it says where it changes and by how much.

Drum kit: the kit against the overheads

At the end of the list of groups is Drum kit: the whole kit, source by source, with the overheads as reference. What the program does, in this order:

1. Moves the overheads once

   The whole pair, by what it takes the hit of the source they carry best (nearly always the snare) to reach them; between 2 and 3.6 ms in a normal kit. The attacks land together without moving the drums against the bass and guitars. If it can't be measured well, it doesn't move them and says so.

2. Decides each source's polarity

   With the overheads moved: if inverted gains at least 1 dB of body, 7 hits out of 10 say so and its lows don't lose, all the mics of that source flip together. The overheads are never inverted.

3. A time tweak per source

   Comes set only if it measures well and also gains body; if not, its row says how many ms its attack asks for, to try them.

• One row per source (kick, snare, each tom, hi-hat…) and, last, the overheads (reference). In each: ON, Ø, Move, Monitor, Listen and Graph. What you do here applies to all the mics of the source at once, on top of what each carries in its group.
• The overheads are at 0 where the program puts them, together with the snare. To the right, later: more depth. With their ON off, they stay as recorded.
• Listen (L): with some L lit, only those rows are heard; with none, the whole kit. Graph puts the row on the graph or takes it off.
• The graph: one waveform, side to side, with every source and the overheads overlaid, each in its colour. Zoom with the wheel down to the milliseconds, drag to move, and double-click a hit to see it close.

Hi-hat, ride and cymbals are rows without a proposal: set by ear (in the highs the measurement can't be trusted). The room and the kit-wide mics don't show in Align.

With its processes

Align is the first thing done to a track, and it's measured with the tracks as they come. But what's summed in the end are the processed tracks. The PROCESSED / UNPROCESSED button changes the window's audio: in PROCESSED, the graphs, waveforms and listening use each mic already processed (everything but Align), with the ms and Ø on top. It comes on PROCESSED in groups where some mic carries an FX Chain. If the mics of the source don't carry the same FX Chain, the window says so.

A mic between kick and snare ("Kicksnare" or similar) is measured with the kick on kick hits and with the snare on snare hits, and only what suits both is proposed. It shows in Align in the kick's group and the snare's, with the same control in both.

A mic without a reliable measurement shows without a proposal: unticked and with the controls at zero, to set by hand if you want.

Phase rotation (bass and kick)

Sometimes moving and inverting isn't enough: bass DI and amp, or kick in and out, can be fine in one zone of the lows and cancel in another, because amp, speaker and mic turn the phase differently at each frequency. There TapeLeader also tries an all-pass filter on the mic being corrected: it doesn't change the level of any frequency, it only turns the phase in the lows (what a Little Labs IBP does). It's offered on kick and bass, only if the sum gains at least 1 dB in some low octave without losing more than 0.5 dB in any other. It's one of the options (the last letter, with φ and its frequency); sometimes it sounds better than the recommended one, sometimes just different, and that's chosen by ear. In the table: D Ø -0.74 ms φ 77 Hz.

9.2 High-pass filter

The cut is measured on each track; it isn't a fixed value per type.

1. What really belongs to the track

   Instruments with notes (vocals, guitars, keys, strings, winds): the lowest note they really play, a low note that repeats, with harmonics on top, not hum or bleed. Bass: that same note (E1, 41.2 Hz, on a four-string); below it is the slow movement of the string as it's plucked, not music. Only on a synth bass do sub-lows count. Keys: their lowest note, unless content is heard further down (low registers, pedals). Kick and toms: the lowest that is heard: the loudest third-octave is put at 85 dB SPL (a loud listen) and the lowest band still above the threshold of hearing (ISO 226:2003; under 20 Hz, Møller and Pedersen 2004) and 10 dB over its gaps is taken. So the tail of the hit's pitch below its fundamental counts, and the air the head pushes (infrasound, which no one hears) doesn't. Snare: where its own sound, on its hits, is at least 3 dB above what it picks up on the kick's hits; below that, it's the kick leaking in. Hi-hat and ride: never below 300 and 200 Hz (they play over kick and snare). Other hits, overheads and rooms: where what sounds only on this track starts, 10 dB over what's in its gaps and within 30 dB of the maximum, for at least 1/6 octave. If it can't be measured, the lowest note the instrument can give.

2. Options A and B

   A (the default): the cut goes where the filter takes 1 dB off that point: with 48 dB/oct zero phase, at 0.77 times its frequency (at 24 dB/oct, 0.59). 1 dB on the lowest note is under the level difference that can be told apart: it cleans without touching what plays. B cuts right at that point (−6 dB there in zero phase): it cleans more, as you would when mixing. An HPF gives no level: it cleans bleed, air and rumble.

3. A ceiling per type, for safety

   Only in case the measurement goes wrong: what's measured on the track rules.

4. Only if there's something to take

   Rumble, stand knocks, wind, bleed. Ticked by default when the note or the floor is measured on the track, on instruments with notes, or when there's real rumble; if not, proposed for you to decide.

5. One cut per source

   All the mics of one source get the same filter, with the lowest cut of the group, and if one mic has an HPF, they all do: with the snare top filtered and the bottom not, their sum loses body. If you take the HPF off one mic by hand, that's respected.

On the HPF card (PROCESS and miniMix): the Freq knob, dB/oct and the options USER · A · B. On hover over each button, the reason, with figures: where the lowest of the track comes from and how it was measured, where it cuts and how much it takes there, how much energy is left under the cut and what it is (infrasound, rumble, kick leaking in, plosives…), which filter and why, and, if it shares the filter with other mics, what it would ask for alone. The same in the table's HPF cell and in INFO. If an option has no room (it would fall under 15 Hz), its button is off and says why. Choosing A or B changes every mic of the source. Moving the knob or the slope goes to USER (and the table cell is filled, like any retouch); USER goes back to the last thing you set by hand. Clicking the active option again goes back to the previous one, as in Align. ↺ brings everything back to the proposal. In Track Presets, Option (A or B) is each type's default; A on all.

A real example (McLovins, option A): kick In and Out → 22 Hz (the lowest heard of the In starts at 28 Hz); snare top and bottom → 96 Hz (its own low end starts at 125 Hz); toms → 55 and 27 Hz; overheads → 29 Hz; bass → 32 Hz (E1); backing vocals → 151 and 134 Hz; guitars → 63 Hz (E2); Hammond → 27 Hz; the KORG, no HPF (nothing to take under it). Each proposal explains the note or floor measured, where it cuts and how much energy is under it; the spectrum of the selected track shows it drawn.

Which filter: zero phase or minimum phase. Zero phase doesn't turn the phase of anything, so doubles, pairs and mics of one source keep summing exactly the same; the price is a slight pre-echo around the cut. It's a Butterworth run forwards and backwards (−6 dB at the cut), applied in blocks with the same result as on the whole file. Minimum phase, like an analogue filter: no pre-echo, but it turns the phase near the cut. In Auto it's chosen per track: zero phase for subsonic cleaning (under 30 Hz), sustained sounds (vocals, guitars, keys: the pre-echo is masked), overheads and room, and kick, snare and toms: they sum with the overheads, and in minimum phase the filter would change their body in that sum and what Align decided (measured on three sessions: on the snare up to 2 dB and about 1 ms; on a tom, even the polarity; on the kick, up to 7 dB in the lows and the polarity). Minimum phase on the other hits with an audible cut (hi-hat, percussion), where the pre-echo would be heard. Slope in Auto: 48 dB/oct zero phase (24 on the kick when the cut is above 30 Hz: up to 9 dB less pre-echo), 18 dB/oct minimum phase.

A tip for the mix. The same holds later in your DAW: if you filter the mics of one instrument, do it on their bus or with the same filter on all; and if you filter the snare or toms with a normal (minimum phase) EQ, listen with the overheads playing, not solo.

There's no preparation LPF. It was removed: nobody uses one to prepare a track. A saved session or preset that had one opens the same, without it. The ultrasonic LPF (9.7) still always runs.

9.3 DC offset

The mean of each channel is measured and, if it's above −60 dBFS, subtracted. That touches neither the phase nor the lows. If there's a slow drift, it warns you; a very low (5 Hz) zero-phase filter can then be chosen for the whole session, in the settings file or with --dc filter.

9.4 Mains hum: subtracted, not filtered

Mains hum is the "mmmm" of 50 Hz (Europe and most of the world) or 60 Hz (North America) and its harmonics, from the building's wiring, ground loops and amps. No gate, no notch: TapeLeader measures the hum and subtracts it.

1. Measure

   The amplitude and phase of each harmonic, in one-second stretches.

2. Learn it in the quiet parts

   Where the music spoils the measurement, the phase is rebuilt from the reliable stretches around it.

3. Synthesise and subtract

   What isn't hum stays intact, notes on those frequencies included (a bass at 100 Hz, say).

Safe: if in some stretch the subtraction would add instead of take, that stretch isn't touched. Measured on real sessions: the hum of an electric guitar went from −18 to −50 dB and that of a bass amp from −16 to −40 dB. The hum is measured over the whole track, also when you listen to a stretch in AFTER: what you hear is what renders. It's the first process on the track, before Align and the filters. If the hum isn't steady enough to model, the track is left alone: a notch would take what plays on those frequencies and, on a hit, leave ringing what it removes. In the settings file (fixes.hum_method, chapter 18) you can choose the classic narrow notches (2 Hz per harmonic) for everything.

9.5 Gate

1. Detector

   The track (mean of its channels) goes through the detection filter, if there is one, and its peak is measured in 1 ms frames. The filter only decides when to open: the audio isn't filtered.

2. Threshold

   If the mic's bleed is measured, between the bleed and its own hits. If not, automatic: the gaps (30th percentile) and what sounds (99th); a third of the way up, at least 8 dB over the gaps and always 12 dB under the signal.

3. State

   It opens on crossing the threshold; it closes when it falls under the threshold minus the hysteresis and the hold has passed.

4. Gain

   Open, 0 dB. Closed, the range (gate) or, as an expander, what the ratio says under the threshold, never past the range. Transitions are ramps of the attack (opening) and release (closing), and the lookahead moves the curve forward so attacks aren't cut.

Where the default values come from. Engineers' guides agree on the principles though almost none give fixed numbers: attack as fast as possible on drums (0–1 ms) and 5 to 15 ms on vocals; short release on kick and snare (50–150 ms) and long on toms so their ring isn't cut; hold around 25–100 ms on vocals and snare; hysteresis around 6 dB; detection filter cutting highs at 200 Hz for the kick, a 200–500 Hz band for the snare and a cut of highs at 400–700 Hz for toms; 2:1 to 4:1 expanders for vocals. TapeLeader's own detection defaults: kick 30–200 Hz, snare 150 Hz–3 kHz, toms 50–600 Hz, hi-hat above 3 kHz. Since this is preparation and not mixing, the range is more cautious than in the guides: on drums, a 2:1 expander that brings things down 6 dB at most, and 4 dB on kick and snare (Mike Senior, in Sound On Sound, suggests a range of about 6 dB on a snare mic for a more natural sound). On drums and percussion, hold, release and the detection filter are set per track from what is measured (chapter 8), and a mic with only soft hits isn't proposed a gate. On kick and snare, the SC (side-chain) gate listens to the other drum (chapter 6).

9.6 Gain

• Each track goes to its type's target with the chosen measure (active RMS, LUFS or peak), never passing its true peak ceiling.
• The ceiling is checked twice: when the gain is worked out, with the original's peak, and at render, with the track already filtered. An HPF can raise a peak by a few dB; if after the filters the track passes, it's lowered just enough, the same for every mic of that source, and the report notes it.
• At most it raises 24 dB.
• The mics of one source get the same gain; so do the two channels of a joined pair (unless you tick "Match L/R").
• It's the same gain on the whole track: the dynamics don't change. The level is worked out on what will come out: a joined pair or a track turned to mono is measured as it will be.

With an FX Chain, two levels. An FX Chain changes the level: a compressor lowers it; a tape or a saturation raises it. And its plugins expect a specific input level: many analogue emulations, 0 VU = −18 dBFS; others, −12 or −20. So: 1. Input level: the gain takes the track to the level the chain asks for. 2. Final level: at the chain's output the whole track is measured again, with the type's measure, and adjusted to the target without passing the ceiling. Mics of one source that go through the same chain get the same final adjustment. With the chain live in the DAWproject there's no final adjustment: you decide the output level in the DAW.

9.7 Ultrasonic low-pass (always)

Every track comes out with a zero-phase low-pass that leaves everything audible intact (up to 20 kHz, deviation under 0.001 dB) and takes what's above: −100 dB from 24 kHz, or from just before the limit at 44.1 and 48 kHz. Many plugins (saturation, tape, coloured compressors) oversample, and ultrasonic energy turns into audible distortion; taken out first, plugins work clean. The filter's ringing lasts a couple of milliseconds and sits above 20 kHz: inaudible.

9.8 Edge fades

If a track doesn't start or end in silence, it gets a 5 ms fade at the edge to avoid clicks.

9.9 Strip Silence

The "Detect Silence" of Cubase, Strip Silence in Pro Tools and Logic. On a long track with long silences (a vocal that comes in now and then, a guitar that only plays the choruses, a tom), the track is left in clips with audio and the rest out. It isn't the gate: the gate brings things down a few dB between hits; this takes out whole stretches, with pre-roll, post-roll and fades.

Breaths aren't cut. The threshold sits close to the noise floor (a breath is well above it); a silence under 1 s is never cut; and on vocals each clip starts 300 ms before and ends 500 ms after.

The threshold it proposes: where there's real silence between pieces, 10 dB over the measured noise floor of the track, never within 20 dB of the loud parts; on a track already edited (digital silence between pieces), 45 dB under the loud parts; on a tom, what's between hits isn't silence but the rest of the kit, so the threshold goes between that bleed and its hits.

Where it comes on: where it cuts at least 5 s and 8 % of the track, on vocals, guitars, bass, keys, strings, winds, loose percussion and effects. On toms, only if the bleed is well under the hits and there are no soft hits the cut would take. On kick, snare, hi-hat, cymbals, overheads and room it stays off: there, the gate. Mics of one source are cut in the same places.

The Strip Silence window

• On top, three boxes: the track, its type, how many clips are left and how much audio; Wave | dB; and the player: ▶, LOOP, BEFORE (the track as it comes) and AFTER (with the silences cut, heard at once).
• The whole track, small, with a box: the stretch shown below. Drag it, stretch it or click outside it; double-click, the whole track.
• The stretch, large: what's cut shows dimmed, each clip between two lines with a green line under it, and the fades drawn. A click plays from there. The wheel zooms; Shift + wheel moves. Wave shows the waveform; dB, the level from 0 to −72 dB, where breaths, bleed and tails show, and the Open and Close lines can be dragged.
• Below: on the left the list of Strip Silence presets (New preset…, Rename…, Delete); in the middle the controls, eight knobs in two rows; on the right the waveform. Monitor out under the box, and why Strip Silence is proposed or not on that track.
• RESET goes back to the proposal, Apply keeps what you see, Cancel leaves it as it was.

What rules on each track, from most to least: the controls you moved on that track; the preset you gave that track; what was measured on it; its type's preset; the defaults.

How it comes out of the render (chosen when rendering): Clips (default): the WAVs come out whole and in the Reaper project and the DAWproject each piece is a clip with its position and fades, so you can stretch any edge. Audio: the cut is written into the WAVs, for dragging them into any DAW. Clips + audio: both. Without a project, it always goes in the audio. It goes after the gate and before the gain and the FX Chain: with a chain that has a tail (a reverb), a cut written in the audio comes before it and the tail is kept.

9.10 Clipper: same peak, more level

What it is. TapeLeader's own soft clipper, with four characters. It needs no plugin and no setting: the program places it on each track with what it measured on its hits.

What it does. The crest of a kick or snare hit lasts 1 to 3 ms. It sets the track's peak, but you hardly hear it as level: what you hear is the body behind it. Clipper rounds only that crest, sample by sample, with no attack or release: the body doesn't notice, nothing pumps and the phase doesn't move. Then it raises the track by what it shaved. The hit goes back to its earlier peak and the body comes up. In the IN / OUT meter: the light bar (the peak), the same; the solid one (the RMS), higher.

Where it comes on. On kick, snare, toms, percussion and overheads, when the track has enough hits, alike, with a crest to shave. Elsewhere it comes off and you turn it on with its button. Mics of one source go together and come up by the same amount.

It's a peak shaver, not a saturator: it gives no tone. For the sound of a particular box, your plugins in an FX Chain.

What goes in and what it does. Two lines, useful too for setting a plugin of yours: Hits −6.8 dBTP · Crest 16 dB · Knee −20.7 dBFS (the level the hits come in at, how far the crest sticks out of the body, and where the curve starts to round) and PEAK +0.0 dB · Body +2.0 dB (what the hit's peak changes and how much its body comes up).

How much it gains. Measured with the kick, snare and toms of a real session: the true peak stays the same or up to 1 dB lower and the loudness goes up between 1.5 and 2.6 LU.

With oversampling. ×4 up to 48 kHz and ×2 at 88.2 and 96 kHz (at 176.4 and 192 kHz there's already room), with linear-phase filters: the track doesn't move in time. It goes after the gain and the ultrasonic LPF and before the FX Chain, and it's written into the WAV.

The Clipper window

Opened with ··· on the Clipper card, by double-clicking the Clipper column title, or from Track Presets. Here the only thing that changes between BEFORE and AFTER is Clipper:

The waveform: BEFORE in the colour of the IN bar and AFTER in that of OUT, half transparent: where AFTER's colour shows is what the body gained; where BEFORE's shows, what was shaved. The wheel zooms down to each sample. ◀ Hit and Hit ▶ jump to the previous or next loud hit at 120 ms on screen. The lines are levels of what goes in: Hits (dotted), Knee (dashed) and Ceiling (solid).

The figures: PEAK (true peak), RMS, LUFS and Crest, of BEFORE, AFTER and their difference, for the stretch or the whole track.

Clipper presets: character, Shave and PEAK MATCH with a name. Seven come with it: Auto (the type's character, Shave Auto), Clean Clip (Hard, Auto), Soft Glue (Soft, 3 dB), Medium Punch (Medium, 3 dB), Tube Open (Tube, 3 dB), Hard Clip (Hard, 5 dB) and Crush (Soft, 7 dB).

9.11 FixEQ: the tone of its instrument, by a broad stroke

What it is. TapeLeader's own EQ that takes each track towards the typical tone of its instrument, so the rough mix sounds decent before mixing. It isn't a mixing EQ: three very wide, musical bands, and a few narrow cuts where something rings.

Towards where. Towards the recommended EQ for its instrument: TapeLeader knows the typical shape of the kick, the snare, the bass, the vocal… (22 in all, measured with pink noise and a little darker in the highs than a mix, because a raw take is) and works out what each track is missing and has too much of. Toms, overheads, room and cymbals go with the whole kit's; the vocal with a low or high voice's, by where it sings. FX, click and the reference mix don't carry FixEQ.

What it measures with. On kick, snare, toms and percussion, on their own hits; elsewhere, where it plays. With several mics of one source, their sum is measured and all carry the same curve: the source goes towards its instrument and each mic keeps its own; since the filter is the same, the alignment between them doesn't change.

What's less than 1 dB from the reference isn't touched, and it hardly boosts below where the track starts or below its HPF (never on cymbals and shakers below 300 Hz).

Resonances are only looked for where each instrument usually has them: the same zones the warnings use (chapter 6). A peak that sticks out 6 dB or more from its octave gets a dynamic band that can bring it down by up to three quarters of what it sticks out (3 to 6 dB), only when it sounds loud. Keys, strings, brass and synths have no zone: nothing is cut, so their character stays.

• Tone: how much of the recommended EQ (100 % is what was worked out; 50 %, half the dB in each band; 0 %, nothing).
• Low and High: dB added to (or taken from) the low and high shelves the program chose, −9 to +9. 0 is the program's. If the program didn't touch the lows (or highs), the shelf appears at 120 Hz (or 5 kHz). They work even with Tone at 0 %. On mics of one source, Low moves on all of them at once; High goes per mic.
• FixReso: a dynamic EQ on the resonances. Each band listens to its frequency and only cuts when it sounds above usual on the track: on a snare or a tom, in the hit and its tail; between hits, nothing. Each mic has its own and listens to its own. The figure is the Range of the deepest; the other goes in proportion. Off on the right and more cut to the left. In grey, "Off", if the track has no resonances.

The level. FixEQ comes out at the level it went in (a Trim measured on its spectrum), so the gain, Clipper and the FX Chain get the track as they expect.

The FixEQ window

Opened with ··· on its card or by double-clicking the FixEQ column title. On top, the track, its type, what it carries and the player: LOOP, BEFORE (without FixEQ), AFTER (with it), GAIN MATCH and ▶. On the left, FixEQ with its four knobs and what was measured; below, the bands with Freq, Gain, Q and phase, and the Trim. In the middle, the graph: on top the spectrum (BEFORE filled, AFTER as a line and, dashed, the recommended EQ) and below the FixEQ curve (±12 dB), with each band marked and the cuts in red. On the right, the IN / OUT meter.

10Per-track settings, A/B and the miniMix

Right click on a track

• Zero-phase high-pass, Gate in the gaps, Strip Silence, Remove hum, Convert to mono, Exclude from render: applied even if the analysis doesn't ask for them. A hand-added high-pass uses the cut (and the filter type) adapted to the track; a gate uses its type's settings.
• Listen A/B… and Align… (on a track with mics of the same source).
• Align: undo the hand-made group, on a track of a group saved by hand earlier. (Making a new group by hand, for mics the names don't relate, will come back inside the Align window.)
• FX Chain: By track type, None, or a specific chain only for those tracks.

A/B listening

A stretch of the track is processed with the whole plan (alignment, filters, hum, gate, gain and FX Chain) and compared with the original. The stretch is the track's loudest part, where the process is heard most. A (original) is matched in loudness to B (with the plan). Both play together in a loop and only one is heard: switching doesn't cut or jump. Space bar to switch.

The miniMix: checking a few tracks together

The miniMix button opens a small desk, with one channel per track you choose, playing over a stretch of the song in a loop. It's where you hear what can't be decided track by track: whether the kit sums with the overheads, whether bass and kick go together, whether a gate shows with everything else playing. It isn't for mixing (that's the Rough mix, after the render) and it's meant for a dozen tracks at most.

[SCREENSHOT: the miniMix with the kit and the bass]

Which tracks go in. In the main table, next to the type, there's a miniMix column with a switch per track. By default: the kicks, snares, hi-hat, toms, overheads and one room (the first in your track order), the bass and the lead vocal. A right click on a track has a miniMix menu: Only these tracks, Add these tracks, Remove these tracks and The ones the program proposes. This switch isn't the render's: a track taken out of the miniMix still renders.

SAVE CHANGES or DISCARD. What you touch in the miniMix is heard and applied at once. SAVE CHANGES keeps everything (it's what will render); DISCARD leaves the session as it was when you opened it. If you close with changes, it asks.

Starting faders. The first time you open the miniMix of a session (all faders at 0), as soon as the stretch is ready the faders set themselves as a first take of a mix: the mics of each instrument against each other (the Out a little under the In, the bottom well under the top, the DI under the amp, a reinforcing sample), the kit built from the overheads (kick and snare reinforcing on their hits, the room and the hi-hat under, the toms a little under the snare and a little more if they have bleed nothing cleans, the second room out), the kit, bass and backing vocals in proportion with the lead vocal, and the master at about −6.5 VU (−24.6 dBFS RMS, the median of the engineer's own mixes). What isn't drums, bass or vocals gets its fader halfway up (−7.4 dB), for you to place. Starting faders (top, next to Proposed section) does it again; if you'd already moved a fader, nothing is touched on its own. It isn't a mix: it's a starting point to retouch by ear. DISCARD undoes it too.

The stretch. Before the render the whole song isn't processed: a half-minute stretch is prepared in seconds. TapeLeader proposes it, straddling the point where the song lifts (usually the end of a verse and the way into the chorus). It doesn't count the song's intro, and if the song never lifts it proposes its fullest part. With the tempo measured, it starts on a beat and lasts whole bars. On top is the whole song with the stretch marked; that drawing is the level of the sum of all the tracks along the song. Drag it, pull its edges (4 s to 2 minutes), click outside to move it there or inside to jump to that point; Proposed section goes back to the program's. It's saved with the session.

The master. On the right. On top, a measurement screen, like a box's: INTEGRATED loudness large (LUFS); below, SHORT (3 s), RMS, PEAK and TRUE PK (red over 0); and MAX, the highest since it was reset, with the OVER light. A click on the screen resets it. Below, MONITOR BOOST: desk keys (0, +6, +8, +10, +12) lit amber. Then its fader, and a VU with a needle, −20 to +3 VU like a real one: 0 VU = −18 dBFS RMS. The pan law (−3 dB or 0 dB) is in Settings ▾. No limiter: if it goes over, you see it and hear it.

With few tracks the channels are wide; as you add, they narrow. With a short window (a laptop), the processes pair up to leave the fader its travel.

What's heard at once and what takes a moment. Fader, pan, solo, mute, polarity and milliseconds are heard as they move. A filter, a gate or an FX Chain makes that track prepare again in the background (a coloured dot on its channel says so) and it comes in on its own, without a click and without stopping the music.

BEFORE / AFTER. AFTER is each track with its whole plan, as it will render. BEFORE is the tracks unprocessed, with the same faders and pan. The space bar plays and stops. Since the render changes each track's level, GAIN MATCH comes on: in BEFORE each track is raised or lowered to the level of its AFTER, so what you compare is what the processes do, not the volume.

The panel of the selected track (below). Full screen, on the left, the track's data, as in the main window's INFO tab, and why each process carries what it carries, one line per process. On the right, its controls, in small cards: the same cards as the PROCESS tab. What you touch here is that track's and is saved with the session. What the program proposed isn't lost: a green dot on each knob marks it (the arc turns green when the knob is there; double-click goes back to it), the title button's tip says it, and the small ↺ in each card's corner brings the whole card back.

• Align: its button puts the alignment on or takes it off (polarity and ms stay in the card). On a mic aligned with another of its source, its options, USER · AUTO · A…, the same as the Align window, to try them playing with the mix. On the overheads, the Move knob. Elsewhere, Delay (±20 ms) with the Ø next to it. ··· opens the Align window on its group.
• DeBleed: Amount (−30 dB to off) and Attack (0–30 ms), on the close mics of kick, snare and toms (chapter 6).
• HPF: the Freq knob, the slope (dB/oct) in one button that says which it carries (12 → 18 → 24 → 48), and the options USER · A · B (A loses 1 dB at the track's lowest, B cuts right there; 9.2).
• Gate: Threshold (bottom, Auto = measured on the track), Range, Hold and Release; on kick and snare, SC and SC Range.
• FixEQ: Tone (%), Low, High and FixReso, and ···, its window (9.11).
• Strip Silence: its button, the track's Strip Silence preset and ···, which opens its window.
• Clipper: its button, the character (Mode) in one button, the Shave knob and ···, which opens the Clipper window.
• FX Chain: its button puts it on and takes it off; the list (its type's, — for none, or a specific one) and, below, the plugins it carries. ··· opens FX Chains.

What stays (if you save): each track's processes as you left them, and the faders and pan set here, which are the ones in the Rough mix and in the DAW project. They don't change the audio written: the tracks render at their working level.

The level after an FX Chain, the ceiling after the filters and the gate's automatic threshold are measured here on the stretch and, at render, on the whole song, so they can differ a little.

11The AI

Without AI and with AI

TapeLeader works completely without AI: offline, without an account and without a model installed. Everything measured in the audio (levels, filters, gate, phase, hum, key) is algorithms that use no AI and can be checked. AI is an expert mode for what's written in human language where the rules fall short: odd track names, the client's notes, each studio's way of working. With it, the app decides better; without it, what's doubtful is marked for you to check.

Privacy: the AI is only sent names, measurements and texts (the notes), never audio. With a local Ollama model, nothing leaves your computer (a model whose name ends in -cloud runs on Ollama's servers).

What it does, one by one

• Review when the session opens. With "When opening a session, review doubtful items automatically and read the notes" (on by default), the AI runs on its own if there are doubtful tracks, unread notes or a mix template with doubtful slots. If it fails, it doesn't get in the way. Also by hand with Classify with AI.
• Types and roles. Decides the type of doubtful tracks and whether one is an alternate take, a printed effect or a guide.
• Pairs and mics of one source, in the cases the measurement doesn't settle.
• Notes. Interprets the Readme and what you paste from the client (chapter 4).
• Review. Which version to use, suspicious tracks and doubtful phase. Everything shows in Warnings unticked, with its reason.
• Preset. Proposes one as a starting point; if you tick it, it's used.
• Texts for the client: a summary of what arrived and what was done, and an email asking for what's missing (files that didn't arrive, clipping with the minute, mixed sample rates, no reference mix or tempo). The draft comes without AI, with only the facts; "Write with AI" writes it better without adding anything that isn't in them.
• "Why?" In the ASK tab of the selected track: it answers with that track's plan and measurements.
• Orders. In the box next to the session title: "take the gate off the toms", "the room in mono", "FX Chain My vocal on every vocal", "tempo 92". The AI can only turn them into actions of the app, and before applying anything it shows you the list to tick.
• FX Chains with your plugins (chapter 12).

Providers

In AI settings… you choose the provider, the model ("Find models" lists the available ones), the API key (or leave it empty and put it in the provider's environment variable, e.g. OPENAI_API_KEY) and: Allow cloud models (untick it to use only local ones), Also check the clear tracks (slower) and When opening a session, review doubtful items automatically and read the notes.

How it makes sure it chooses well

• Rules first. The AI only acts where the rules hesitate, or when you call it.
• Closed answers. It always answers in a strict format, and the values can only come from the app's lists: the types, your tracks, your presets, your installed plugins, the possible actions. Anything not in the list is thrown away. It can always answer "I don't know", and that's better than guessing.
• It never decides figures. Levels, cuts, thresholds and delays are measured from the audio.
• Checked. If the answer isn't valid, it tries once more; if it fails again, the rules stay.
• Cached. Asking the same twice gives the same answer and costs nothing.
• Visible. What the AI decides carries the "AI" mark and its reason. What removes something is proposed unticked. What you correct by hand always wins.
• With Ollama, if it isn't running, TapeLeader starts it; if the model doesn't fit the graphics card, it retries on the processor. Before asking, it checks the AI answers; if not, "AI not available" shows at the bottom and everything goes on without it.

12FX Chains and plugins

Opened with FX Chains in the bar. The installed plugins (folders and scan, set up once) have their own window: Settings ▾ → Installed plugins….

FX Chains

An FX Chain is an ordered list of your VST3 plugins applied to the track types you choose in Track Presets (or to single tracks). Filters and gate don't go in FX Chains: they're set per track type.

• Add… opens a search box: type a few letters and the plugins that contain them in their name, maker or kind remain ("ua 1176", "pulsar eq", "reverb").
• New chain…, Rename…, Delete, Move up, Move down, Remove, and each plugin's box to skip it without removing it. Four plugins at most per chain (one saved with more still plays, but Add… is off). Chains are stored in ~\.tapeleader\chains\.
• Input level of the FX Chain: what level tracks go in at. Auto: drums by their hits (recommended): drums and percussion go in by their hits, everything else at its type's level, so the same chain serves a kick and a bass. Or The track type's, A custom one: (in dBFS of active RMS: the 0 VU its plugins expect) or By the hits:, in dBTP: the typical loud hit of any track reaches the chain at the level the chain keeps, so it works the same on all. Two snares with the same RMS can have their hits several dB apart. At the output it's always measured again and left at the final level (9.6).
• Hits in: the level this chain's hits are coming in at, with the track you're hearing. From this track: when the chain sounds as you want with that track, press it; the chain keeps that level and the hits of any other track come in there. It switches to By the hits and Gain in goes back to 0: what you hear doesn't change.
• Use on…: the chain you're hearing, for this track, for its track type, or as a copy only for this track.
• Oversampling of the selected plugin: No, ×2 or ×4, plugin by plugin. Tail: seconds of silence added at the end for reverb or delay tails.
• Suggest with AI… (with an AI set up and plugins scanned): the AI proposes an FX Chain per track type only with your installed plugins. They're created without settings: you set them.

Listening to the chain and setting its plugins

Listening is of the whole chain, in real time. A (BEFORE) is the track already processed, as it reaches the chain; B (AFTER), after all the active plugins. The stretch is chosen on the small waveform, 3 to 40 seconds.

• Open the plugin window: the plugin's own window, always above TapeLeader. Several at once, closed in any order. On top of each, a small TapeLeader bar with Bypass and OS ×1 ×2 ×4.
• Turning a plugin on or off reloads nothing: every plugin of the chain is loaded and an off one is just skipped. No window closes, no setting is lost, it's heard at once.
• Save plugin settings: saves how all the chain's plugins are now, as .vstpreset in its folder (any VST3 host can load them), with each plugin's controls next to it. If a plugin doesn't take its saved settings (after an update, say), the chain loads with it at its defaults and says so.

Listening and analyser

• Gain in and Monitor out: thick sliders in 1 dB steps. Gain in is how much over its input level the track goes into the chain; it's saved with the chain and the render goes in the same way. Monitor out is for listening only.
• GAIN MATCH (B at A's level): in real time the engine itself does it while it plays, comparing a second and a half of input and output, so it readjusts itself when you switch a plugin off or move a control.
• CURVE: what the chain does to each frequency, 20 Hz to 20 kHz. Measured on its own with pink noise through the chain (not heard) at its input level: half a second for compressors to settle and a second of measurement. A sweep was used before, and with a compressor the curve lied: the gain reduction changed along the way and showed up as EQ. What the curve doesn't show is how a compressor moves in time: that's seen in the meters and heard.
• Spectrum, in bars over the curve: the bars are the chain's output; the input is a thin line through them. Where the line is above a bar the chain takes; inside it, the chain adds. Bars of 1/3, 1/6 or 1/12 octave, fall speed and peak hold to choose.
• DRIVE: the peaks going in, in dB against where the chain starts to clip (Clip) or compress (Threshold), measured by SWEET SPOT. Green under the sweet zone, amber inside, red past the clip. GR: what the chain lowers the level while it plays, measured against the same chain clean with that track, so EQ doesn't count, only what presses.
• VU with two needles: BEFORE (thin) and AFTER (thick), real VU ballistics, scale to +3 VU. 0 VU = Auto by default: as on a box, 0 VU is where the chain works at its sweet spot. Or fixed, from −20 to −12 dBFS.

SWEET SPOT

The SWEET SPOT button next to CURVE: what the chain does by how hard it's driven, and at what level it works at its sweet spot. An EQ curve can't show a clip or a compression; this can. It works for any plugin or chain: TapeLeader doesn't look at names, it sends through the chain, as you set it, tones from −60 to +6 dBFS, a loud tone followed by a soft one, and the stretch of the selected track, and looks at how it answers. One or two seconds.

• The graph: across, your Gain in (−48 to +24 dB). On top, the colour the body takes at each Gain in (THD, 0.001 % to 100 %); below, what the hits lose (dB the crest of a typical hit comes down). The red line Gain in is where you have it. Clip (or Threshold) is the Gain in at which hits start to clip (or compress); the green zone, SWEET SPOT, where to put it.
• What the chain is, in one sentence: linear (EQ, nothing to propose), colour without clipping, colour with clipping, clipper, or compressor, limiter or tape.
• Four figures of the stretch with the chain as it is, against the same chain 30 dB lower: Hits, Loudness, Distortion (on a compressor, Body) and Tone.
• What it proposes, and the Sweet spot you want for this chain (saved with it). For a colour, the body's distortion (by default 1 %, what a Fatso's 0 VU marks); with clipping, also the hits 1 dB under where it starts to clip. For a compressor, limiter or tape that presses, what the hits come down (by default 1 dB). For a clipper, what's clipped off them (by default 1 dB). It never lets the hits in above −1 dBTP.
• Warnings: memory (it takes time to come back after a loud hit), aliasing, a chain that rises a lot clean, and peaks that grow.
• The SWEET SPOT button by the Gain in takes the chain's Gain in to its sweet spot with the track you're hearing; the plugins' controls aren't touched, and it holds for every track that uses the chain. It says how far it'll move it; ✓ if it's there already. Auto, next to it: every time SWEET SPOT measures, the Gain in goes there on its own.

Emphasis

The EMPHASIS box: one or two bands TapeLeader raises before the chain and lowers exactly after. That band goes harder into what saturates and saturates sooner (a snare's crack at 4–5 kHz, the body at 200 Hz…). With negative Gain, the opposite: the band is protected while the rest saturates. It's the pre-emphasis of tape and vinyl, and works with any saturator, with or without EQ, because the cut is always the exact mirror of the boost.

• Freq · Q · Gain per band. At 0 dB a band does nothing: that's how it comes; TapeLeader doesn't propose tone.
• Around: the whole FX Chain (default) or a single plugin, for when more than one presses.
• BAND MATCH (on by default): when it saturates the band loses some level in the loud parts; on, TapeLeader measures how much on average and gives it back in the cut, so the average tone stays and only the loud hits of that band are pressed.
• Its bells show in green on CURVE. What it does on the track, against the same chain without Emphasis, shows under SWEET SPOT.
• It holds for every track with this chain; each band's Gain can be changed for a type or for a track: the track wins, then the type, then the chain. Live in the DAWproject, a chain with Emphasis goes printed.

Mix

Under Tail: how much of the whole chain you hear, 0 % (the track as it goes in) to 100 %. The processed signal is matched in level to the dry one before mixing (it changes the colour, not the volume) and the dry is delayed by what the chain really takes (measured), so there's no comb filter. What can't be aligned is the phase the plugins turn: on CURVE a dashed curve says how the mix comes out and, if an octave loses more than 1.5 dB, it's shown in a warning colour.

Macros

In progress: built and parked while they mature; the Mix of a macro isn't tested yet. The MACROS box: four knobs and two switches, each moving at once whichever controls you want in the chain's plugins, each with its range, like the macros of Blue Cat's PatchWork. A macro can also move the chain's Gain in, and a switch can bypass a plugin.

• AUTO: the macros the program proposes by reading the names of each plugin's controls: Drive, Mix, Tone and Output, and as switches the bypass of the first and last plugin. Proposing them changes nothing of what sounds.
• LEARN: touch a macro and move a control in a plugin's window: it's assigned to that macro, without jumps.
• ··· opens the list of what each macro moves: name, and for each control Min and Max, ⇄ to invert the range and × to remove it; Add control…, Gain in, Chain mix, Bypass… and LEARN.
• Values per type (Track Presets) or per track (PROCESS and the miniMix): the track wins, then the type, then the chain. At render each track goes through the chain with its own values; live in the DAWproject, each track gets its own preset.

How listening works: in real time

• A and B always line up: the engine compensates the plugins' latency, so input and output, and the IN and OUT bars, are the same instant of the music.
• The curve is measured apart by a second set of plugins with the same settings as the ones playing: each plugin is loaded twice, one sounds and the other measures, so the test signal is never heard and never disturbs the state of what plays.
• The oversampling per plugin goes in real time too: what you hear is what renders.
• If real time can't start or stops (the card won't open, a plugin won't allow it), TapeLeader switches to listening by passes on its own and says so.

Oversampling per plugin

Plugins that saturate (tape, valves, clippers) create harmonics above what fits at 44.1 or 48 kHz; if they don't oversample inside, those harmonics come back as false tones in the audible range (aliasing). With oversampling, TapeLeader raises the rate before that plugin and brings it back down after. Chosen plugin by plugin (×2 or ×4), never above 192 kHz. Up and down don't touch the audible range and the result stays in the same place in time. Measured with a simple saturator with no oversampling of its own (two tones, 5 and 6 kHz): the strongest false tone comes down more than 30 dB with ×4 (the app's own test).

Plugins run in a separate process: if one hangs, TapeLeader doesn't. It says which FX Chain and track it was, and leaves that track unprocessed.

Installed plugins

• VST3 folders: by default C:\Program Files\Common Files\VST3. Add folder… and Remove.
• Scan: tests each plugin in a separate process. If one hangs or crashes, TapeLeader doesn't. The scan gives maker, category, declared and real latency, and whether it works. VST2 (.dll) can't be loaded. Plugins that can't load stay in the list with their reason. Instruments aren't tested. While it scans the button is Stop: what was tested is kept, and next time it carries on where it was.

13Output names

Double-click the title of the Output ▸ column to open Output names. On top is the format: pick one or write yours with these markers, and both the window's list and the table's Output column show at once how each track will be called.

A name by hand for one track. Double-click its Output cell, or its name in the window's list, and write it (without .wav). It wins over the format: if you change the format later, that track keeps its name. Empty, it goes back to the format's; Clear hand-typed names does it for all. Two files can't have the same name: a "(2)" is added, and characters a file doesn't allow are replaced. The markers are written in Spanish in every language.

Built-in formats: {n}_{nombre} (default), {n} {plantilla}, {n} {limpio}, {n} {familia} - {limpio}, {n}_{tipo}_{limpio} and {sesion}_{n}_{limpio}. For versions, write it yourself: {n} {tipo} {version} → 01 Kick In, 06 Tom.

Learn from examples…: write how you'd call two or three tracks and TapeLeader looks for the format that gives exactly those names. It first tries every sensible combination; if none fits and you have AI, it asks the AI, and its answer is only accepted if it gives exactly your examples.

Your mix template

If you always mix with the same DAW template ("Kick In, Kick Out, Snare Top, Snare Bottom, Hat, Tom 1…"), TapeLeader can number and name the tracks in that order, so you can drag the WAVs onto your template (Pro Tools, Logic, Cubase…). The numbers run on without gaps: a slot this session doesn't fill isn't skipped. The Reaper project and the DAWproject come out in that order too.

Number by type numbers by family (drums, bass, guitars, keys, strings, winds, vocals, FX) and keeps the original order inside each.

In the same window, Mix template → Manage…: New from a pasted list… (one track per line, in order), Import from Reaper / DAWproject / .txt…, then check the type of each slot (guessed from its name) and save.

How it places tracks: by type and mic position (In/Out, Top/Bottom, DI/Amp, L/R) and by number (Tom 1, Tom 2). Several tracks for one slot ("Gtr" with three guitars) go together. With AI, what doesn't fit by type is asked to the AI, which can only choose a slot from the list or "none". Tracks without a slot go at the end, by family. Pairs joined in stereo get a common name with {limpio} (Room Left + Room Right → Room).

14Output format and WAV metadata

• Sample rate: the original, or converted to 44.1, 48, 88.2 or 96 kHz (soxr at its highest quality). If the session mixes rates, everything comes out at one.
• Bits: 32-bit float (default): never clips. 24 and 16 bit: with TPDF dither; if anything passed 0 dBFS it's clipped and you're warned.

Metadata

• BWF (bext, version 2): time stamp (by default the session starts at 0 in the DAW) and the measured loudness of the output (integrated LUFS, range, true peak, momentary and short-term maxima).
• iXML: project, a note with the original name, each channel's name, instrument and title (ASWG), and TapeLeader's data (type, group, colour and every operation).
• LIST/INFO: title, project and comment (the ones the Windows Explorer shows).
• ACID (optional, off by default): tempo. Cubase can switch to musical mode on import.
• Files over 4 GB switch to RF64 on their own.

15Rendering

The Render button (top right, or Ctrl+R) opens a window with everything that needs deciding to write the tracks:

• Destination: the folder. A subfolder with the session's name is created inside.
• Audio format: rate and bits (chapter 14).
• In addition to the WAVs: Reaper project, DAWproject, HTML report and click track: a tick on each beat measured in the session, in the Reference folder and muted in the project. Only offered if the tempo in use is the measured one (or its half or double) and steady. All ticks are equal: where the bar's "one" is can't be measured.
• Non-destructive project:
  • Gate as automation (editable, not printed) (on by default): the WAVs come out without the gate, and the gate goes as volume automation: Volume (pre-FX) in Reaper and track volume in the DAWproject. On tracks with a printed FX Chain the gate is printed (it goes before the chain). Cubase doesn't import DAWproject automation yet: for Cubase, print the gate.
  • Live chains in the DAWproject (not printed): the plugins of each chain, with their settings, open in Bitwig and Studio One (in Cubase, check it in your version) instead of processing the audio. A chain with Emphasis or Mix below 100 % is printed anyway. The presets go in the Presets folder, for Reaper too.
  • A copy of the originals in a muted folder: sample for sample when the rate doesn't change, in an "Originals" folder that goes muted into the project.
  • Strip Silence: Clips (default), Audio, or Clips + audio (9.9).

On top, a summary of what will come out: how many tracks, how many with processes and the preset in use. Everything is remembered for next time.

[SCREENSHOT: the render window]

What it writes

<output folder>\<session>\
├── Audio\               the WAV of each track
├── Reference\           the reference mix, if any, and the click, if asked for
├── Originals\           a copy of the originals, if asked for
├── Presets\             presets of the live FX Chains, if asked for
├── <session>.rpp         Reaper project
├── <session>.dawproject  Bitwig, Studio One, Cubase 14+, VST Live…
├── <session> rough mix.wav   the Rough mix audio, if you export it
└── Report\              report.html, report.json, plan.yaml and mix.json

• Reaper project: tracks with name and colour, folders by family, tempo and its changes, the notes' markers, faders and pan, the key in the project notes and the reference mix, muted. WAVs are linked by relative paths: you can move the whole folder.
• DAWproject: the same except the key (markers, tempo changes, gate automation and, if asked, the live FX Chains) in the open format Bitwig, Studio One and Cubase 14+ read (in Cubase 15: File > Import > DAWproject; folders, names, colours, faders and clips arrive, but Cubase doesn't import the automation or the tempo track yet). The WAVs are linked by their full path; if you move the render folder, the DAW will ask where they are.
  If Cubase does nothing on import (not even a message): it's a Cubase issue with the Windows temporary folder when Windows writes it abbreviated, with ~1 (it happens if a folder in the path has more than 8 letters, a long user name for instance). It doesn't import any DAWproject, not even Cubase's own. Fix: in Windows, "Edit environment variables for your account", set TEMP and TMP to a folder whose path has names of 8 letters or fewer and no spaces (E:\TempWin, say), and sign out.
• HTML report: the key and the markers, and a table with each output (type, channels, gain, LUFS, true peak and waveform) and what was done with it and why. report.json and plan.yaml: the same, for machines.

When it finishes, TapeLeader offers to open the Rough mix or the output folder.

Rough mix

The quick mix of what you just rendered: "in five minutes you make a mix and see how much signal reaches the master". It opens at the end of the render or with the Rough mix button, to the right of Render, which is lit only when the session has a render. Closing it with changes offers Send to the DAW project.

[SCREENSHOT: the Rough mix]

It's the miniMix's desk without processes or panel: one channel per rendered track (fader, pan, solo, mute and meter) and the master with its readings and monitor boost. With more than ten tracks the channels go in two rows (split between two families: the kit isn't split). The whole WAVs play; a click on the song on top jumps there. Faders and pan start as you left them in the miniMix.

• On a track with the gate as automation, the WAV doesn't carry the gate: here it's applied, so it sounds as in the DAW. A track with its FX Chain live plays without it.
• Export the audio… writes the mix as it sounds (with the monitor boost) to <session> rough mix.wav. If it passes 0 dBFS it warns before writing and offers to bring down what's over.
• Send to the DAW project writes the render's .rpp and .dawproject again with these faders, pan and mute (and the master fader), and adds the mix sheet to the report: a table with each track's fader and pan, for any DAW. The boost doesn't go. If the project was saved from the DAW after the render, it warns before writing over it.
• The mix saves itself next to the render and, if it's the open session, in it too: the next render comes out with those faders.

Fader dB are exact in any DAW. Pan doesn't sound identical in all, because each compensates the centre its own way: choose your DAW's pan law in Settings.

16Saved sessions

• Autosave: everything you decide in a session (boxes, types, per-track processes, FX Chains, preset and changes, names, output folder, and the miniMix: faders, pan, mute, the stretch and what you retouched on each track) saves itself every time you change something. Open the session again and it's as you left it.
• What's remembered of the boxes is what you changed, not the whole list: if you unticked an HPF it stays unticked; what you didn't touch follows what the program proposes today.
• Save session… creates a .tapeleader file with all that, to archive or pass on. Open session… opens the original library and session and recovers what was decided. If the session's files were renamed or reordered, TapeLeader matches them again by name.
• On start, the last library and session open.
• The dot next to each session in the list: empty, never opened; filled, opened with saved decisions; green, rendered. State and autosaves live in the state folder next to the cache.

17Themes, look and language

The main window always opens full screen, on the monitor it was on when you last closed the program. All the others can be moved and resized, and each reopens where and at the size you gave it, the next day too.

In Settings ▾ → Theme:

The change is immediate and remembered. Each theme brings its typeface, included in the app. In all of them, each track's type carries its instrument's icon in its family's colour (blue the drums, purple the bass, amber the guitars, teal the keys, green the strings, gold the winds, pink the vocals; in the DAW projects the drums go red): a different icon for each of the 47 types. Number columns use fixed-width figures.

Sound device

Settings ▾ → Sound device…: the device (ASIO too) and the buffer. Asked the first time TapeLeader opens.

Language

In Settings ▾ → Language: Spanish and English; it applies after closing and reopening. The first time, Windows' language is used (English if it's neither).

In both languages, studio words stay in English, the way they're said and written on plugins: gate, threshold, attack, hold, release, lookahead, range, ratio, HPF, gain, ceiling, headroom, oversampling, buffer, hum, bleed, click, fades, clipping, DC offset, loudness, sweep, sample rate, bit depth, bypass.

Still in Spanish only: the HTML report, most of the command line help, and for now the reasons of Align and of the classification, and the AI's answers. The texts for the client come in Spanish or English (choose in their window). The theme and AI provider names in the menus are still in Spanish too. Warnings of a session analysed earlier stay in the language it was analysed in until it's analysed again.

18The settings file

The app keeps its settings in a YAML file: tapeleader-windows.yaml in the TapeLeader folder when started with TapeLeader.bat, otherwise %USERPROFILE%\.tapeleader\settings.yaml. Your own presets live in ~\.tapeleader\level_presets\, FX Chains in ~\.tapeleader\chains\ and mix templates in ~\.tapeleader\templates\. Almost everything is changed from the interface; these are the most useful settings by hand:

cache_dir: D:\TapeLeader\cache     # extractions, analysis and state
auto_analyze: true                # analyse when opening a session without analysis
ui_language: auto                 # es | en | auto (Windows' language)
ui_theme: pizarra                 # estudio | rack | consola | aire | tapeleader | tapeleader_oscuro | pizarra | grafito | claro | papel | sistema

library:
  max_depth: 4                    # folder levels explored
  seven_zip_path: null            # path to 7z.exe if not in the usual place

render:
  target_sample_rate: null        # null = the session's; 44100 | 48000 | 88200 | 96000
  bit_depth: 32                   # 32 (float) | 24 | 16
  dither: true
  dc_mode: auto                   # auto | mean | filter | off
  fade_ms: 5
  zero_origin: true               # the session starts at 0 in the DAW
  write_info: true                # LIST/INFO
  write_acid: false               # ACID tempo

fixes:
  level_preset: "Preparar: 0 VU (-18 dBFS RMS)"   # the preset's internal (Spanish) name
  link_groups: true               # same gain for the mics of one source
  gain_max_boost_db: 24
  hpf_order: 4                    # 4 = 48 dB/oct, 2 = 24, 1 = 12
  hum_method: subtract            # subtract | notch
  hpf_option: A                   # A (−1 dB at the lowest) | B (cut right there)

naming:
  template: "{n}_{nombre}"
  sort_by_type: false
  layout: null                    # name of your mix template

user_aliases: {}                  # learnt words: "gonzo": kick

llm:
  provider: off                   # off | ollama | openai_compatible | openai | anthropic | gemini
  model: ""
  allow_cloud: true
  auto: true                      # review what's doubtful when a session opens
  api_key_env: null               # or the name of a variable holding the key
  review_all: false               # also check the clear tracks (slower)

plugins:
  folders: []                     # extra VST3 folders
  tail_s: 3

export:
  reaper: true
  dawproject: true
  html_report: true
  folders_by_family: true
  reference_track: true
  gate_as_automation: true        # gate as automation, not printed
  live_chains: false              # FX Chains live in the DAWproject
  originals: false                # copy of the originals in a muted folder
  click: false                    # click track with the measured tempo
  silence_output: clips           # clips | audio | both

audio:
  buffer: 1024                    # 64 to 2048 samples

mixer:
  pan_law_db: -3                  # pan law of mono tracks: -3 or 0 (dB at centre)
  section_seconds: 30             # length of the stretch the miniMix proposes
  gain_match: true                # in BEFORE, each track at the level of its AFTER
  start_faders: true              # starting faders the first time

19Command line

The main steps (scan, analyse, classify, render, verify) can also be run from the console, to automate or to batch many sessions. The console doesn't read what you decided in the window (per-track changes, Align options, miniMix faders): it builds the plan fresh. DeBleed, FixEQ, the HPF option and the Strip Silence output mode aren't in it yet.

All except version and verify accept -s settings.yaml and --cache folder. Without -s, the console uses the factory settings, not yours.

tapeleader render "D:\Sessions\Song.zip" -o D:\renders -s tapeleader-windows.yaml --gain --levels "Balance: pop/rock" --sr 48000 --bits 24 --names "{n} {limpio}" --by-type

20Troubleshooting

The session shows "not analysed", or there are no warnings.

If it's an unextracted archive, press Analyze. If not, it's usually that a new version measures something more and the old analysis no longer applies: the session analyses itself when opened.

"AI: not responding", or no Ollama models show.

Check Ollama is installed and has a model downloaded. TapeLeader tries to start it if it isn't running. If you use another address or port, set it in AI settings….

"The model doesn't fit in memory" with Ollama.

TapeLeader retries on the processor. If it doesn't fit either, choose a smaller model or close programs using the graphics card.

The measured tempo is half or double what I expected.

Measuring can't tell them apart: choose Half or Double on the tempo button.

No key, or "neither major nor minor".

Without enough material with notes (only drums, say) no key is given. "Neither clearly major nor minor" is an answer on purpose: in blues and rock on fifths both thirds sound almost equal. You can set it yourself.

Windows asks permission for "Python" while setting a plugin.

Plugins load in a separate process that Windows sees as Python. If it's the Windows Firewall, it's usually the plugin checking its licence online: you can allow it (or not, if the plugin works offline).

A plugin's window doesn't open.

FX Chains says why under the listening area.

A plugin hangs or closes.

Plugins run in a separate process with a time limit: TapeLeader keeps going and warns you. Take that plugin out of the chain or scan again.

A RAR or a split archive won't open.

Install 7-Zip or give its path in the settings file.

The render warns of clipping at 24 or 16 bit.

Some track passes 0 dBFS after processing (with "No level changes" and a chain that raises level, say). Render at 32-bit float or check that track's gain.

The miniMix or the Rough mix don't play.

They say why, top right (the card busy with another program, a different sample rate with ASIO…). The device is the one in Settings ▾ → Sound device….

PDF notes aren't read.

Reading PDFs needs the pypdf module: run test_on_windows.bat again, which installs it.

Something to send to support.

The logs cache\plugin_host.log (plugins) and cache\mixer.log (miniMix and Rough mix).

A track doesn't come into the miniMix (its channel keeps the coloured dot).

It couldn't be prepared: the reason shows top right. Most often an FX Chain with a plugin that fails; take the chain off in its panel or check it in FX Chains.

Align proposes nothing for a group of mics.

It only proposes if the sum improves without any band getting worse. The reason shows in the list ("already aligned", "the sum doesn't improve"…).

21Glossary

0 VU

Analogue reference level; in digital it's often aligned to −18 dBFS RMS (EBU), −20 in SMPTE practice.

dBFS

Decibels relative to digital full scale (0 dBFS).

dBTP / true peak

The real peak between samples, measured with oversampling.

LUFS

Unit of perceived loudness (ITU BS.1770).

Active RMS

Average level measured only where the track plays.

Clipper

TapeLeader's own soft clipper: shaves the crest of the hits and raises the body (9.10).

Shave

The dB Clipper takes off the crest of the track's typical loud hit.

PEAK MATCH

Raising the track by what was shaved: the hit goes back to its peak and the body comes up.

Crest

How far the crest of a hit sticks out of the body behind it, in dB.

DELTA

In the Clipper window, hearing and seeing only what the process changes: AFTER minus BEFORE.

Knee · Ceiling

The level where a curve starts to round, and what the output never passes.

Hit peak

The true peak of a track's typical loud hit (not of one louder stray hit).

Zero phase

A filter that delays no frequency: it doesn't change the phase.

Minimum phase

A filter like an analogue one: no pre-echo, but it turns the phase near the cut.

Pre-ringing

A small ripple before an attack produced by zero- or linear-phase filters.

Bleed

What a mic picks up from an instrument that isn't its own.

Fundamental

The lowest of the frequencies that rule a sound: the "note" of a kick, snare or tom.

Range · Tessitura

From the lowest to the highest note a voice sings, and the zone of it where it sings almost all the time.

Sibilance

The S's and other hissing consonants of a voice, 3.5 to 12 kHz.

Plosive

The air blast of a "p" or "b" on the mic: a low pop.

Harshness

A narrow resonance in the upper mids that hurts when the track is turned up.

Noise floor

What a track has when its instrument is silent: bleed, room, hiss.

GCC-PHAT

A cross-correlation method for measuring delays between mics.

Coherence

How alike the waveforms of two tracks are, 0 to 1: high on two mics of one source, near zero on two different takes.

Doubled takes

Two performances of the same part, panned apart. Not joined into stereo.

miniMix

The small desk for checking a few tracks together before the render. Not for mixing.

Rough mix

The quick mix of faders and pan over the rendered tracks.

Pan law

How much a mono track drops at the centre so it doesn't sound louder than at a side: −3 or 0 dB depending on the DAW.

Monitor boost

Raising the master only to hear and meter the mix louder; it doesn't go to the DAW.

Chromagram

How much each of the 12 notes sounds at each moment; used for the key.

Strip Silence

Cutting a track's silences: it's left in clips with audio and the rest out.

Clip

Each piece of audio on a track in the DAW (event in Cubase, item in Reaper, region in Logic).

Pre-roll · post-roll

What a clip starts before the track opens and goes on after it closes.

TPDF dither

Minimal noise added when reducing bits to avoid quantisation distortion.

BWF · bext · iXML

Standard metadata inside professional WAVs.

DAWproject

An open project format between DAWs (Bitwig, PreSonus).

DeBleed

TapeLeader's process that takes the other drums out of each close mic, also inside the hit, learning from the session.

SC

Side-chain: the gate of a kick or snare listens to the other drum to close on its bleed.

FixEQ · FixReso

TapeLeader's broad EQ that moves each track towards the tone of its instrument; FixReso, its narrow dynamic cuts on resonances.

SWEET SPOT

The measure of what an FX Chain does at every input level, and the Gain in it proposes.

Emphasis · BAND MATCH

A band raised before an FX Chain and lowered exactly after it; BAND MATCH gives back what the band loses on average when it saturates.

Macros

Four knobs and two switches that move several plugin controls of an FX Chain at once (in progress).

Starting faders

The miniMix faders set as a first take of a mix the first time it opens.

Automation

A control that moves on its own along the song in the DAW: the gate goes as volume automation.

Mains hum

The 50 or 60 Hz of the mains and its harmonics, picked up by cables and amps.

Gate

Brings a track down between its own sounds; as an expander, gently and by a ratio.

22Coming next

What's planned, roughly in order. Nothing here is in the app yet.

• FixEQ references recorded per mic: you EQ each mic yourself in the FixEQ window and save it; with enough of them, a reference of your own for each mic (Kick In, Kick Out, Snare Top…).
• Levels by instrument: a default that leaves each track where a good engineer would (drums by their peaks, sustained tracks measured the way a VU reads them), and more presets: a harder one and one towards −12.
• Gate options (USER · AUTO · A · B), as in Align, to try alternatives playing.
• The bleed matrix of the whole kit, so the level gate and what INFO says leaks into each drum can be trusted with hi-hat and toms too; and with it, a drum's own hits told apart when another drum comes in almost as loud.
• The gate's Range measured on each track, the automatic threshold shown as a figure instead of "Auto", and the ghost notes counted in INFO.
• Clipper wherever there are hits, not by type.
• Hear what you see in FX Chains: each thing SWEET SPOT and CURVE show, playable with one click and level-matched (what's added, the harmonics, the aliasing, what the hits lose), and a compare window over an FX Chain (BEFORE, AFTER, DELTA) with how much it shaves on the hits, and a warning if the sum of two mics gets worse after it.
• Trim and SWEET SPOT per plugin, for two drive plugins in a row, each at its sweet spot.
• Macros, finished.
• Recognising the instrument by its audio when the name says nothing.
• More versions chosen by the audio: today it does it on toms (Rack, Floor); still to come on a kick without "In" or "Out", a snare without "Top" or "Bottom" and a bass without "DI" or "Amp".
• MIDI and spikes from kick and snare, to trigger samples, gates or compressors.
• A vocal leveller: a gain per phrase, before the compressor.
• Automatic de-esser and de-plosive: today it warns, with the moment and the frequency.
• More warnings on vocals: breaths, mouth clicks, headphone click.
• Resonances that keep ringing after the hit, measuring how each one dies away.
• The gate drawn over the waveform (where it opens and closes).
• Strip Silence in Pro Tools and Logic as clips (today, written in the audio there): needs AAF.
• Moving the kit against the overheads from the miniMix with a proposal (today it proposes only the polarity; the milliseconds, by hand).
• The harmonic spectrum before and after Clipper, several tracks at once in its window, and more marked characters if 2–3 dB feel too little.
• Tempo map for sessions without a click, and the bar's "one".
• Export to AAF and a Pro Tools session.
• A version for macOS.
• Bleed between instruments of a band recorded in one room (today, only the drums).
• Phase rotation also on snare top/bottom and several mics on one amp (today, kick and bass).
• A Windows installer.
• This manual as a PDF, with screenshots.