AI Models
Use AI models when you need to recognize sounds such as extended techniques, breath, key clicks, or percussive gestures.
Use a Model
ONNXMODEL "flute.onnx"
ONNXDESCRIPTORS mfcc zcr centroid spread
BPM 60
PTECH pizz C4 1
sendto sample [start pizz_echo]
UTECH jet-whistle 2
sendto noise_texture [open]
Rules:
- event labels must match labels used during training;
- descriptor order must match the training order;
- model paths are relative to the score file.
Prepare Training Data
Flute/
pizz/
pizz_C4_01.wav
pizz_D4_01.wav
jet-whistle/
jet_01.wav
jet_02.wav
Each subfolder is a class label. Use short, isolated, representative recordings.
Train and Use
- Record examples.
- Organize them by label.
- Extract descriptors.
- Train an ONNX model.
- Load it with
ONNXMODELandONNXDESCRIPTORS.
See also: Audio Analysis, Advanced Configuration, AI Model Reference.