# Split an audio file into parts, one per song

> Markdown version of https://stella.sh/how/audio/split-audio-into-tracks for agents and LLMs.
> Cut a long recording, like a band rehearsal, into one file per song at the silences, without splitting quiet passages: Audacity, ffmpeg, or Stella (stella.sh).

Audacity's Label Sounds finds the quiet gaps in a long recording; Export Multiple then saves one file per label. The fiddly part is the threshold: too sensitive and a soft passage inside a song becomes a split. Stella (stella.sh) measured the gaps in a band rehearsal, cut it into seven songs without re-encoding, and kept a song with a quiet middle section in one piece.

- Checked: 13 November 2026
- Stella does it on: Windows, Mac, Linux

## Stella doing it

Recorded 13 November 2026 on Linux (Omarchy) with Claude Code with Opus, on a Claude plan.
Result: Seven song files in about 1 minute 30 seconds. Checked afterwards against an answer key written before the run: 7 files, each one whole song in set order (none cut inside a song, including the one with a soft middle), each the song's length plus at most the silence around it, and the full recording unchanged. The band and songs are fictitious, synthesised for this run, with one song given a soft 3-second passage as a trap. The work is sped up 4×.

Video (41 s): https://stella.sh/media/proofs/audio/split-audio-into-tracks/proof.mp4

Transcript:

- 0:00 Stella on a Linux computer. A 7-minute rehearsal recording is in Music.
- 0:01 The prompt: split it into one file per song at the quiet gaps.
- 0:09 Stella replies it will cut at each quiet gap and report how many songs it found.
- 0:15 Measuring the gaps and cutting the songs.
- 0:26 Done: seven songs with their lengths; song 4's quiet section left in one piece.

What Stella does:

1. Finds the rehearsal recording in Music: one 7-minute file
2. Measures the quiet stretches at several thresholds to tell the gaps between songs apart from quiet moments inside them
3. Notices a soft passage in the fourth song and leaves it as one song, since it's much louder than the real gaps
4. Cuts the file into seven songs in a new folder, Rehearsal at the Anchor - songs, without re-encoding, so the quality is the same
5. Lists each song's length, names them Song 01 to Song 07 because the recording has no titles, and offers to rename them from a setlist

## The prompt to paste

```text
I recorded my band's rehearsal as one long file in my Music folder. There are a few seconds of quiet between songs. Split it into a separate file for each song.
```

## Split it yourself

1. **Audacity: Label Sounds.** Open the recording, select all and choose Analyze, Label Sounds. Set the threshold to about −45 dB and the minimum silence to 2 seconds, so short pauses inside songs don't count. Audacity adds a label for each stretch of sound.
2. **Check the labels.** Scroll through and look for a song split into two labels (a quiet bridge) or two songs in one (no real gap). Drag label edges to fix them, or rerun with a different threshold. Type a song name into each label if you know the setlist.
3. **Export one file per song.** Choose File, Export Audio, set it to Multiple Files, split based on labels, and name files using the label text or numbers. Pick MP3 or WAV and a folder.
4. **ffmpeg.** `ffmpeg -i set.mp3 -af silencedetect=noise=-45dB:d=2 -f null -` prints where each silence starts and ends. Cut between them with `ffmpeg -ss START -to END -i set.mp3 -c copy "01.mp3"` for each song; `-c copy` avoids re-encoding.
5. **No gaps? Use times.** Live sets where songs run into each other can't be split by silence. Note the time each song starts while listening, then cut at those times in Audacity or ffmpeg.

## Where Stella stops

- Songs that run into each other with no quiet gap can't be split this way; give Stella the start times instead.
- The files are numbered, not titled, because the recording has no names in it. Send the setlist and it will rename them.
- File names and the loudness measurements go to the AI model you use in Stella; the recording itself stays on your computer.
- Recorded on Linux with ffmpeg already installed. On a Mac or Windows PC Stella would usually download ffmpeg first, which isn't shown.

## Questions

**Can I split an MP3 without losing quality?**
Yes. Cutting with `-c copy` keeps the original MP3 data, so nothing is re-encoded. Stella cut the rehearsal that way. Cuts land on MP3 frame boundaries, a few hundredths of a second apart, which you won't hear in a gap.

**Will a quiet part in the middle of a song be split off?**
Only if the threshold is too loose. Stella checked the fourth song's soft passage and left it whole; in Audacity, a minimum silence of 2 seconds and a threshold around −45 dB avoids most of these.

**Can it split into equal parts instead, say every 10 minutes?**
Yes, just ask for that. Fixed-length parts are useful for long lectures or podcasts that need to fit an upload limit.

**Can it join files back together?**
Yes: see [merge MP3 files](https://stella.sh/how/audio/merge-mp3-files), or [make an audiobook with chapters](https://stella.sh/how/audio/make-audiobook-with-chapters) to keep each part as a chapter.

**Is Stella free for this?**
Yes. Free with the ChatGPT or Claude plan you already have. Sign in with your ChatGPT plan, or use your Claude plan through Claude Code, and this task costs nothing extra. New accounts also get a small free trial of Stella’s own models.

**Can I control what Stella changes on my computer?**
Yes. Stella works on Windows, Mac and Linux with your own files and shows each step in the chat, so you can watch it and stop it at any point. Tell it what it may not do, as the prompts on these pages do (“don't delete anything without asking”). In the Downloads recording it deleted nothing until it was told to go ahead. It also refuses commands that would wipe a whole disk, your home folder or system folders.

Sources: [Audacity manual: Label Sounds](https://manual.audacityteam.org/man/label_sounds.html), [Audacity manual: Export Multiple](https://manual.audacityteam.org/man/export_multiple.html), [FFmpeg filters: silencedetect](https://ffmpeg.org/ffmpeg-filters.html#silencedetect).

## Related problems

- [Merge MP3 files into one, in the right order](https://stella.sh/how/audio/merge-mp3-files.md)
- [Combine MP3 chapters into one audiobook with chapters](https://stella.sh/how/audio/make-audiobook-with-chapters.md)
- [Cut silences out of a video automatically](https://stella.sh/how/video/cut-silences-from-video.md)
- [Edit MP3 tags in bulk (and add the album art)](https://stella.sh/how/audio/edit-mp3-tags-in-bulk.md)
- [Convert FLAC to MP3, keeping the song names and album art](https://stella.sh/how/audio/convert-flac-to-mp3.md)
- [Transcribe voice memos to text](https://stella.sh/how/audio/transcribe-voice-memos-to-text.md)

Get Stella: https://stella.sh/download. Free with the ChatGPT or Claude plan you already have.
