Best tool for open-source voice cloning

Posted by Safe_Ad_8485@reddit | LocalLLaMA | View on Reddit | 5 comments

I have been trying to do voice cloning for some time for my personal project, experimented with Coqui XTTS v2 and F5-TTS, the results were not so great,
trying tuning via the parameters no luck.
https://github.com/coqui-ai/TTS
https://github.com/swivid/f5-tts

want to know the open-source tool which is best for voice cloning ?

[-]

FinBenton@reddit

Omnivoice gives me the best results.

[-]

kurunku@reddit

I've worked extensively with F5-TTS and for my work it has been on par with 11labs and Fish audio if only using English. Also, voxtream2 was released recently and is promising.

[-]

Uriziel01@reddit

I've had the best results with VoxCPM 2 (using the text transcription no the automatic one)

[-]

Safe_Ad_8485@reddit (OP)

thanks, going to try that :)

[-]

Uriziel01@reddit

Also if you are on NVIDIA there is a `Ultimate-TTS-Studio` for unified one-click `Kokoro, KittenTTS, Higgs audio, Chatterbox, Fish-Speech, F5 & index-tts & indextts2` setup, in my tests VoxCPM is still better, but if you just want to quickly be able to test a bunch of solutions and check what works for you this is a very fast way to do it in minutes of setup instead of hours (there is a Pinokio script here https://github.com/pinokiofactory/ultimate-tts-studio). Good luck :)