☰
Google Doc
☾
Google Doc
Google Doc
edit. 28.08.26
▸
- (no longer necessary) Fork of UVR GUI and How to install - support for AMD and Intel GPUs appeared (works only for VR and MDX architectures), Besides W11, also W10 confirmed working, MDX achieves speeds of i5-4460s using 6700 XT, while for VR, speeds are v. fast and comparable to CUDA, so CPU processing might be slower in VR, but for MDX you might want to stick with the official UVR5 GUI.
The best models
for specific stems
SDR leaderboard & explanation
▸
Bleedness and fullness leaderboard Python evaluation script by jarredou (prob. mirror), Torch version (with Bas Curtiz), used on Quality Checker
50 models sorted by SDR
UVR5 GUI (MDX, VR, Demucs 2-4 and UVR team models)
Manual ensemble
UVR’s VR architecture models
First vocal models trained by UVR for MDX-Net arch:
9.703 model is UVR-MDX-NET 1, UVR-MDX-NET 2 is UVR_MDXNET_2_9682, NET 3 is 9662, all trained at 14.7kHz
Demucs 3
Demucs 4 (+ Colab) (4, 6 stem)
Gsep (now GAudio) (2, 4, 5, 6 stem, karaoke)
dango.ai
music.ai
Iterative Ensemble Colab by IntroC
MDX23 by ZFTurbo (jarredou fork) - 2, 4 stems
KaraFan by Captain FLAM
Ripple/Capcut/SAMI-Bytedance/Volcengine/BS-RoFormer (2-4 stem)
▸
>Drumsep - single percussion instruments separation
Mel-Roformer MVSEP drumsep models
SCNet MVSEP drumsep models
(newer) MDX23C 5 stem drumsep by jarredou
(older) MDX23C 6 stem drumsep by jarredou/Aufr33
BS-Roformer MVSEP 4 stems
Older drumsep by Inagoy
Moises.ai drumsep
FactorSynth
Regroover
UnMixingStation
LarsNet
VirtualDJ 2023/Stems 2.0 (kick, hi-hat)
RipX DeepAudio (-||-) (6 stems [piano, guitar])
Spectralayers 10
USS-Bytedance (any; esp. SFX)
Zero Shot (any sample; esp. instruments)
SAM-Audio
▸
AudioSep
Medley Vox (different vocalists)
real-time
▸
__Sources of FLACs for the best quality for separation process__
Dolby Atmos ripping
360 Reality Audio FAQ
▸
___AI mastering services___
___Best quality on YouTube for your audio uploads____
▸
___Best quality from YouTube and Soundcloud - how to squeeze out the most from the music taken from YT for separation___
Repository of stems/multitracks from music to create your own dataset
List of cloud services with a lot of space
or for temporary storage
AI-killing tracks - difficult ones to get instrumentals (or vocals) - a lot of e.g. vocal (or instrumental) leftovers in current models
Training models guides
Index
·
Google Doc ↗
Google Doc
›
Google Doc
First vocal models trained by UVR for MDX-Net arch:
Page 6 of 28 ·
Edit this page in Google Docs ↗
← UVR’s VR architecture models
9.703 model is UVR-MDX-NET 1, UVR-MDX-NET 2 is UVR_MDXNET_2_9682, NET 3 is 9662, all trained at 14.7kHz →