Audio Analysis
Music Information Retrieval (MIR)
Music information retrieval for each track: tempo (BPM), musical key, genre, mood, and emotional valence (sad…happy) and arousal (calm…energetic) on a -1 to 1 scale.
When to use it
Tag a music library, build playlists by tempo or mood, check what is playing on a stream, or add music features to a recommendation model.
In a pipeline
Audio File InputMusic Information Retrieval (MIR)Export CSV
Settings
| Setting | Type | Required | Default or hint |
|---|---|---|---|
| Genre and mood (uses an AudioSet model) | checkbox | — | true |
| Model (Hugging Face id) | text | — | MIT/ast-finetuned-audioset-10-10-0.4593 |
| Run on | select | — | auto |
| Time limit (seconds) | number | — | 900 |
Example
Tagging two tracks
| File Name | Tempo (BPM) | Key | Genre | Valence |
|---|---|---|---|---|
| track-01.mp3 | 124.1 | A minor | House | 0.42 |
| track-02.flac | 72.5 | D major | Classical | 0.18 |
Worth knowing
- Tempo and key are measured from the signal itself (onset autocorrelation and Krumhansl key profiles) and need no model.
- Genre and mood come from the AudioSet model's music classes; untick them for a fast, model-free run.
- Tempo can come out at half or double the felt beat for some styles.
- Models run on the server in FlowDoo's Python and are downloaded once, on first use, into the data folder. A GPU is used when the server has one.