Session Clip Export

Exports Session View audio clips to WAV files — each rendered at the current song tempo for exactly one loop length, peak-normalized to −1 dBFS, written as 24-bit PCM. Works on the whole session grid or just the clips you have selected, and can optionally split each exported clip into stems.

Menu entries

Right-click on…	Entry	Does
a scene header	Export Session Clips to WAV	every session audio clip in the set
a scene header	Export Session Clips + Stems	same, plus demucs stems per clip
selected clip slot(s)	Export Selected Clips to WAV	just the selected clips
selected clip slot(s)	Export Selected Clips + Stems	same, plus stems

Multi-select works the way you'd expect: click/shift-click/cmd-click slots in the session grid, right-click the selection, pick an Export entry.

Output lands on the Desktop in Session Clips <tempo>bpm <timestamp>/.

File naming: the clip's name, numbered on duplicates — Bass.wav, Bass 2.wav, … Stems sit alongside as Bass - drums.wav, Bass - bass.wav, Bass - vocals.wav, Bass - other.wav. Unnamed clips become Untitled.wav, …

How it works

renderPreFxAudio only renders the arrangement, so for each clip the extension:

1. Recreates the clip briefly on its own track's arrangement, 8+ beats past existing content, playback starting at the clip's loop start.
2. Renders exactly loopEnd − loopStart beats (marker range for non-looping clips) pre-FX — warping is applied at the song tempo; the track's device chain is not baked in.
3. Deletes the temporary clip — the set is left exactly as it was.
4. Decodes, peak-normalizes to −1 dBFS, writes 24-bit PCM.
5. (Stems) runs demucs on the normalized clip, so the four stems sum back to the exported file, and copies them next to it.

Limits: audio clips only — MIDI session clips are skipped and counted in the log (the SDK renders audio tracks only). Unwarped clips render over their marker range at native speed.

Stems prerequisite

Stem separation shells out to demucs; install it once with pipx install demucs (or pip3 install demucs). First run downloads the model (~80 MB) and separation runs on CPU — expect tens of seconds per clip. If demucs isn't found, the export still runs and the log says what to install.

Install in Live

npm install
npm run package        # → Session-Clip-Export-0.0.1.ablx

Then in Live: Settings → Extensions → drop Session-Clip-Export-0.0.1.ablx onto the Extensions page. That's the whole install.

For development instead: enable Developer Mode (Settings → Extensions) and run npm start. Logs (per-file gain, stem counts, skips, errors) go to ~/Library/Preferences/Ableton/Live x.x.x/ExtensionHost.txt.

Verify

1. Small set: 2 audio tracks, looped session clips of different lengths (1/2/4 bars), a duplicate clip name, one MIDI clip.
2. Scene-header export → WAV count = audio clip count; MIDI counted as skipped in the log.
3. A 4-bar clip at 120 bpm is exactly 8.000 s (ffprobe), peak −1.0 dBFS (ffmpeg -af astats).
4. Duplicates arrive as Name.wav + Name 2.wav.
5. Select 2 of 5 clips → Export Selected → only those 2 exported.
6. With demucs installed, a + Stems export adds 4 stem files per clip.
7. The arrangement is unchanged afterwards (temp clips deleted).