The step before the cover
A Jammable alternative for training your own RVC model
Jammable is a cover-making site: you pick a voice, give it a song, and it returns a finished cover. NiceVois does not make covers. It trains an RVC model from audio you supply and gives you the .pth and .index, which you then use in whichever cover tool you prefer.
If you searched for an alternative because you wanted a custom voice rather than one from a catalogue, training is the part you were missing.
Two different jobs
| Step | Hosted cover site | NiceVois |
|---|---|---|
| Provides a voice catalogue | Yes | No |
| Trains a model from your audio | Limited or not offered | Yes |
| Gives you the model file | Usually not | Yes: .pth, .index, ZIP |
| Separates vocals and remixes | Yes | No |
| Produces a finished cover | Yes | No |
How the two fit together
They are complementary rather than competing. Train the voice here, then take the files into a cover workflow: separate the vocal from the instrumental, convert the vocal with your model, and place it back over the backing track. Our cover guide walks through that, and the after-training guide covers loading the files.
Why train your own? A catalogue voice is whatever the catalogue has. A model you train is a specific voice — your own, a client's, or a collaborator's who agreed to it — and the file keeps working regardless of what any one site does later.
What the beta accepts
Up to 15 minutes of clean, single-speaker audio and up to 100 epochs, in WAV, MP3, FLAC, M4A, or OGG. Finished models remain downloadable for seven days unless you delete them earlier.
Only voices you may use
Because you supply the audio, you carry the permission. Use your own voice or one you have clear consent to train — see the acceptable use policy.
Train it from your permitted audio and download the portable files.
Train my RVC model