CoolFace
Modelpublic

smank/htdemucs-onnx

sourceHugging Facemitupdated 3mo agoView on Hugging Face
0likes
Model Card

HTDemucs — ONNX conversions

ONNX conversions of the Demucs v4 (Hybrid Transformer) family, packaged for SonicDemixer (on-device stem separation for macOS). All graphs share one contract: input mix [1, 2, segmentSamples] float32 planar waveform → output sources [1, numStems, 2, segmentSamples], opset 17, STFT/iSTFT embedded in the graph (no complex dtypes).

FileStemsNotes
htdemucs.onnxdrums, bass, other, vocalsbaseline 4-stem
htdemucs_6s.onnx+ guitar, piano6-source variant
htdemucs_ft.onnxdrums, bass, other, vocalsthe fine-tuned bag's four per-source models fused into ONE graph — each submodel contributes its specialty stem (mapping verified against the bag weights). Higher quality, ~4× the compute.

Integrity: files are LFS-tracked; the LFS OID is the SHA-256 that consuming apps verify after download.

License & attribution

MIT. These are derivative conversions of:

  • —Demucs / HTDemucs — code and pretrained weights © Meta Platforms (FAIR), MIT. https://github.com/facebookresearch/demucs
  • —ONNX-exportable HTDemucs + conv-based STFT/iSTFT — Mixxx Development Team demucs fork (GSoC 2025, Anmol Mishra), MIT. https://github.com/mixxxdj/demucs

Conversion scripts: tools/model-conversion/ in the SonicDemixer repository.