songs_only preset¶
Extract only the singing/music portions from a recording. Removes speech, talking, and silence.
What it keeps¶
- Music segments (instrumental sections)
- Singing segments (vocal melody)
- With
--demix: only pure singing (vocals separated from instruments)
Usage¶
# Basic
praisonai-editor edit concert.mp3 \
--preset songs_only \
--detector ensemble \
-v
# Best quality — with Demucs stem separation
praisonai-editor edit concert.mp3 \
--preset songs_only \
--detector ensemble \
--demix \
--primary-zone \
-v
How content detection works¶
flowchart LR
A[Input audio] --> B[FFmpeg stats\ndetector]
A --> C[Demucs stem\nseparation]
B --> D[Ensemble\ndecision]
C --> D
D --> E{Block type?}
E -->|music| F[✅ Keep]
E -->|singing| F
E -->|speech| G[❌ Remove]
E -->|silence| G
F --> H[Render output]
G --> H
With --demix and --primary-zone¶
flowchart TD
A[Input] --> B[Demucs separates\nvocals + instruments]
B --> C[Classify each segment]
C --> D[Find singing zones]
D --> E[Identify PRIMARY zone\nlargest singing block]
E --> F[Trim to primary zone]
F --> G[Output: just the concert]