search

inagoy / drumsep

Hybrid Demucs model for drum separation
MIT License
89 stars 6 forks source link

ASK for UVR5 #4

Open ghost opened 6 months ago

ghost commented 6 months ago

modelo_final.th and YAML

Where can I download it now?

deton24 commented 6 months ago

https://drive.google.com/file/d/1S79T3XlPFosbhXgVO8h3GeBJSu43Sk-O/view

https://drive.google.com/file/d/1LJ_C4h-hXAJEMaP7lbPTpcrVH-i_VKjm/view?usp=drivesdk https://drive.google.com/file/d/1S79T3XlPFosbhXgVO8h3GeBJSu43Sk-O/view

sob., 20 kwi 2024, 07:19 użytkownik danny69boc @.***> napisał:

modelo_final.th and YAML

Where can I download it now?

— Reply to this email directly, view it on GitHub https://github.com/inagoy/drumsep/issues/4, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIJ3EHH3X7447E34RZ4XVRLY6H3HLAVCNFSM6AAAAABGQGNCUWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGI2TIMZYHAYTENA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

ghost commented 6 months ago

Thanks for the great model! I would love to create my own learning model like this model. Do you think it is possible to make a learning model that extracts only these sounds? So far I've had a lot of failures. https://soundcloud.com/nicole-268639085/test-file

deton24 commented 6 months ago

There's a plan in the community for a newer model with a bigger dataset in the future. The least I can reveal without jinxing it and promising much.

pon., 22 kwi 2024, 19:49 użytkownik danny69boc @.***> napisał:

Thanks for the great model! I would love to create my own learning model like this model. Do you think it is possible to make a learning model that extracts only these sounds? So far I've had a lot of failures. https://soundcloud.com/nicole-268639085/test-file

— Reply to this email directly, view it on GitHub https://github.com/inagoy/drumsep/issues/4#issuecomment-2070416302, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIJ3EHCHPMV3I3ZNYOYERODY6VETBAVCNFSM6AAAAABGQGNCUWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANZQGQYTMMZQGI . You are receiving this because you commented.Message ID: @.***>

deton24 commented 4 months ago

The new drumsep model by jarredou and Aufr33 is out: https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c/edit#heading=h.2u19k7ty9b00

Works with UVR. Colab attached. Better SDR than the original Drumsep. Expanded dataset and better model architecture.

Still, I'd recommend using only on (e.g. separated) drums