Made a test sample comparing the two Collin models using rvc v2. The first one is the one trained by me with v2 48k and inference using Replay (the extraction was also done using replay for fairness and any issues are not the models fault, but the extraction)

Heres the new model.

Replay is the RVC client that works as both an extractor for audio and inferencing.

A better Easy GUI without training (since its being worked on)

Its not a colab at the moment, just a program. I will ask though.

They said they have no plans for a colab version.

No, its supposed to be a replacement for RVC clients using that service.

Don't spoil his ideas.

Correct