the vault


<aside>

If you want to keep up to date with the latest updates on audio machine-learning models, you can find at this google docs link a file that’s frequently updated. The newest information is on top, the file is hundreds of pages, but don’t need to read it all.

</aside>


Machine-learning audio processing

news database

This is the most comprehensive document that covers the latest advances in machine-learning audio processing (stem extraction, cleaning etc.).

📂 docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c

MVSep-MDX23 music separation model

stem splitter

Stem splitting tool. Probably the best out there. It won the 3rd place at the Sound demixing challenge 2023, qualifying a bit lower than the proprietary models from Facebook and TikTok (which are not available for public use).

MVSep-MDX23 detailed use instructions

🟢 https://github.com/jarredou/MVSEP-MDX23-Colab_v2/

Ultimate Vocal Remover

stem splitter

One of the best suite of stem splitting models available. Better than iZotope RX or any other mainstream software that can do stem separation (Logic, SpectraLayers). Also very good for removing the reverb out of sounds (especially vocals). For recordings where there is too much room sound, this may be the solution.

🟢 ultimatevocalremover.com

Untitled-1.webp

Rubin AI Replay

voice cloning