Alyean/recordside-models

Practice Library models

Files the Practice Library app downloads the first time you use a feature that needs them. You don't need to download anything from here yourself.

Stem separation models (ONNX)

Demucs models by Meta AI Research, converted to ONNX (fp16 weights) so the app can run them without Python. Weights and code are MIT-licensed; see LICENSE-demucs.txt and https://github.com/facebookresearch/demucs.

File	Demucs model	Used for
htdemucs_6s.onnx	htdemucs_6s	Fast (built into the app)
htdemucs_ft_vocals.onnx, htdemucs_ft_drums.onnx, htdemucs_ft_bass.onnx	htdemucs_ft	High Quality
htdemucs.onnx, hdemucs_mmi.onnx, mdx_extra_0..3.onnx	htdemucs, hdemucs_mmi, mdx_extra	Bass Ensemble

models.json lists every file with its size and SHA-256; the app checks both before it uses a download.

Soundfonts (soundfonts/)

Optional General MIDI soundfonts for the tab synth (Settings > Audio).

• GeneralUser-GS.sf2: GeneralUser GS v2.0 by S. Christian Collins. License in soundfonts/GeneralUser-GS-LICENSE.txt. https://github.com/mrbumpy409/GeneralUser-GS
• FluidR3_GM.sf2: Fluid (R3) General MIDI by Frank Wen. MIT license in soundfonts/FluidR3_GM-LICENSE.txt.