New Mel-RoFormer vocal removal model
Introducing the new version of Mel-RoFormer, which can really be called the best vocal removal model in the world. This is a huge step forward from the previous version! The new model removes vocals less aggressively, reminiscent of the sound of the old UVR models, but it does it in a much better way.
And for premium users, "correct_phase / de-noise" post-processing is available to make the music sound more natural.
In the spectrograms below, you can see how drastically different the results are.
Previous model:
New model:
