Uhh...excuse me, but, there is a bug(?) that causes infinite loading...?
And even...even after more than 500 seconds...it was still like THAT!
@RuneLightLovely This Space shares the same CPU across all users, so one long training job can block everyone else. I've added safeguards to prevent this (2h audio cap, 8h timeout, cancel button) so this should be fixed now. If it happens again, you can always duplicate the Space for private use: https://huggingface.co/spaces/Luminia/rvc-beatrice-voice-conversion?duplicate=true
Training Status: queue: 2/2 | 4143.4s
@RuneLightLovely Fixed. Added: 5-min timeout & retry on stalled downloads. If a download gets stuck, it now retries automatically instead of blocking forever.
That looks normal, Preprocessing isn't just removing silence. it runs two neural networks on CPU:
- ContentVec (378MB model) - extracts voice embeddings from each chunk.
- RMVPE (181MB model) - extracts pitch/F0 from each chunk.
Just need to wait, keep in mind, More audio = more chunks = more iterations per epoch = longer per epoch.


