Rehearsal Recording Splitter

Automatically split long rehearsal recordings into individual song files.

Detects song boundaries by analyzing audio energy levels — music is loud, chatter and silence between songs is quiet. No manual interaction needed for typical recordings.

Requirements

• Python 3
• numpy (pip install numpy)
• ffmpeg on PATH (brew install ffmpeg on macOS)

Usage

# Basic — auto-detects everything
python3 split_rehearsal.py rehearsal_2026-03-15.wav

# See what it detects without splitting (dry run)
python3 split_rehearsal.py rehearsal.wav --dry-run --verbose

# Tweak parameters
python3 split_rehearsal.py rehearsal.wav \
    --min-song-length 90 \
    --min-gap-length 5 \
    --format mp3 \
    --prefix "rehearsal_03" \
    --output-dir ./songs/

Output files are written to <input>_songs/ by default.

How it works

1. Loads the audio via ffmpeg (supports WAV, MP3, FLAC, and anything else ffmpeg can read)
2. Computes RMS energy in 0.5s windows — music is loud, chatter/silence is quiet
3. Auto-detects a threshold between "music" and "not music" using the energy distribution
4. Finds song boundaries — contiguous loud regions, merging brief quiet moments (e.g. a drum break mid-song won't cause a false split)
5. Filters out segments shorter than --min-song-length (default 60s) to discard short chatter sections
6. Splits using ffmpeg into separate files with optional padding

Options

Option	Default	Description
--output-dir	<input>_songs/	Output directory
--format	same as input	Output format: wav, mp3, or flac
--min-song-length	60	Minimum song length in seconds; shorter segments are discarded
--min-gap-length	3	Minimum gap in seconds to count as a song boundary
--threshold	auto	Energy threshold (0-1). Auto-detected if not set
--padding	0.5	Seconds of padding to keep before/after each song
--prefix	song	Filename prefix for output files
--dry-run	off	Detect and print song timestamps without writing files
--verbose	off	Print detailed analysis info

Tips for blues-rock rehearsal recordings

• If songs get split mid-song (e.g. during a quiet breakdown), increase --min-gap-length to 5-8 seconds.
• If chatter gets included in songs, lower --threshold. Use --dry-run -v first to see the auto-detected value, then set it manually a bit lower.
• Zoom H5 recordings (typically 24-bit/48kHz WAV) are handled natively.
• For long jam sessions, try --min-song-length 120 to avoid counting short sound checks as songs.