Record several passes, choose the best phrases and combine them into a new performance. Recorded audio takes are separate from source-history snapshots and live loop captures; Save ideas, compare versions and record takes explains where to find each kind.
Recording availability depends on your host and audio input. The workflow below is available in desktop Studio; an unavailable recording control reports the host's limitation.
Review and use a take#
Track → Recording brings input/monitoring, recording controls and recorded takes into one place. Preview a take, then Use take as new track to create a one-shot sample phrase with its complete duration. It joins the selected scene; the original recording remains available.
Tick two or more passes of one batch and choose Comp with…: Volta writes a comp that splits their punch window between them (pass 1 plays the first part, pass 2 the next, at whole beats where the window allows), joins the seams with 12 ms crossfades, and plays it on a new track at the bars you recorded — one undo step — then opens the comp lane on it so you can move the seams and swap passes. Run to hear it. The track joins the selected scene when that scene is a written list; for a computed or imported scene the notice says to add it yourself.
voiceComp = comp [
take "voice.t1" 0c 0.5c,
take "voice.t2" 0.5c 1c,
] |> fades 12ms
Loops this song has captured are listed under Loops. Keep keys as take saves the loop's current audio — overdubs included, since that is what you heard — as an ordinary take, keys.k1, then keys.k2: kept beside the song, reloaded on open and offered by the comp lane.
On the Timeline (Arrange), each recorded pass is drawn where it was played, in a TAKES strip under the lanes: translucent, labelled with its name, and its hover names the song version it was played against (played against v12, or the source's first characters when no history version matches). Passes of one window looped in place stack in rows. The clips are read-only.
Capture a live performance#
Choose Capture performance from start in the recording menu. The song restarts; perform scene launches and control changes, then stop recording to review it. Save arrangement as new song creates a finite score containing accepted note gates, actual track-set changes and macro-base movement. Imported source modules are saved as snapshots alongside the native song, or inside the browser document. The original song is unchanged. Failed saves retain the take. Save capture as new timeline song is the other way out, and the one that does not pad: it writes the placements at the boundaries they actually had (Write a song as a timeline § "Capture performance"), so a part that entered at bar 6.5 enters at bar 6.5 and one that rested for three bars rests for three bars. It also carries a Provenance: a take that overran the recorder's slot table, crossed a transport discontinuity or ran through a hot swap cannot claim exact replay, and the saved file says so at its head.
The older Save arrangement as new song still pads its final scene to the next supported scene boundary, because a scene score has no finer grid to land on; note timing is retained either way, and track changes between representable interior scene boundaries keep their original event log.
An automation gesture goes into this same log — one recorder, not a second one. Each point carries the engine sample timestamp beside the musical beat, the destination, the value and the gesture boundary, and the raw take is kept, so the smoothing a Commit applies is reversible. Which gestures are recorded is the destination's automation mode and arm state (Draw a change, and record one § Modes): an unarmed gesture is an audition and records nothing.
This explicit recording is independent of the short display-history buffers. It has a fixed 262,144-event capacity. Overflow, source changes and seeks, loop jumps, clock resyncs and live tempo changes prevent a misleading replay; the original source and event log remain saveable. Imported/composed track definitions, sliding voices, randomly chosen sample variants and unsupported live controls explain why they need an audio print. Instruments and effects remain live in a captured score; this is not a frozen audio render or a copy of external media/plugin state. Save original source and event log keeps the captured evidence when score conversion is unavailable.
Metronome, count-in and tap tempo#
The transport's click control turns on a metronome with an accented bar start. The click is for monitoring: it stays out of recordings, bounces and exports. Choose a count-in of one, two or four bars to hear the beat before the song starts. Enabling count-in also enables the click.
Use tap to set a tempo between 30 and 300 bpm; a pause longer than 2.5 seconds starts a fresh phrase. The transport and its readout follow at once, and 2 s after the last tap the song's tempo = is written (whole bpm unless it has decimals; one Undo step). A tempo map or computed tempo is left alone: the tap plays until the next Run, and a toast says why.
Punch points and recorded passes#
The transport's punch control places a bar-aligned, one-bar recording window at the playhead. Clicking it again clears that window.
Open the recording menu's Recorded audio takes controls. Choose a name, 1–64 passes, overdub mode and punch beats. A punch window is required, and the loop must be at least as long as that window. Arm recording to capture one take per pass; arm again to disarm. Progress shows completed and total passes.
Takes get names you can type. Leave the name empty and a batch is named after the track whose Recording page you armed it from, else your input device (MacBook Pro Microphone → macbook-pro), else take; a typed name wins. The first batch is voice.t1, voice.t2, …; the next batch of the same name is voice2.t1, … (mic-2-2.t1 when the name already ends in a digit). A saved loop is keys.l1, keys.l2, …; a kept loop keys.k1. A name is never reused, even after its file is deleted, and never replaces a sample the bank already holds. Recordings made before this naming keep their names (rec-18f…-4242.t1, keys-loop-18f….t1) and keep loading.
Stereo inputs retain their stereo image; identical left and right channels are stored as mono. Overdub adds each pass to its predecessor and creates a new numbered take, preserving the earlier recording. Latency compensation uses reported input and output latency. If the audio device cannot report comparable timing, capture remains uncompensated.
Find and keep your recordings#
Desktop Studio saves numbered WAVs in the song's takes folder as 32-bit float audio. Each recording batch receives the next unused name, and existing recordings are never overwritten. If saving fails, the take remains available in the current session and Studio reports the error.
History → Recorded audio lists saved takes, links matching source versions and copies sample references. Takes also appear in the Library and reload when you reopen their song. A take cannot replace an existing sample-bank name; a collision produces a warning. See Save ideas, compare versions and record takes for recording history.
Combine phrases with comp#
This example requires the named recordings to be available:
lead = comp [
take "rec.t3" 0c 4c,
take "rec.t1" 4c 8c,
] |> fades 12ms
vox = patch { source = sample lead, gain = vel }
take source from to selects a range from a recording, sample name or prepared sample handle. The bounds must be constant durations. comp joins those ranges end to end in list order; the positions refer to the source recordings, not to the resulting performance. Its result is a sample handle that you can play, chop, stretch or use in another comp.
fades adds equal-power crossfades centered on the joins, using audio outside each selected range where available. Crossfades never change the comp's length, and are limited to half of the shorter neighboring slice. If no extra audio exists beyond a cut, the seam fades through silence.
A missing recording or a range extending past its end contributes silence of the declared duration, preserving the timing of later phrases. Empty comps, empty or inverted ranges, and nonconstant bounds are errors.
A comp may join takes played against different versions of the song — that is a technique, not a mistake. When it does, Studio says so once (compleadjoins takes played against 2 versions of the song …) and builds the comp anyway.
Edit a comp visually#
Open the comp dock on a cell containing an editable comp [take …] list. Each region shows the waveform for its source range, sized by its duration.
| Gesture | Result |
|---|---|
| Drag a region's left or right edge | Trim its source range |
| Drag its body | Slip the source range while preserving duration |
| Select a region, then choose a take chip | Switch that region's recording |
| Scrub the fade value | Change crossfade duration |
| Click a region | Audition its selected range |
Edits update the source code and retain the duration units you used (c is one bar of the song's meter, as the compiler reads it). One drag is one undo step. The source picker includes recorded takes (passes, saved loops and kept loops) and sources already used by the comp, including bounce and generated-sample handles.
A region whose take was played against a different version of the song than the one sounding now is hatched, with a red cap, and its hover says so; the header counts them (1 played against an earlier version).