4 ms·
This is already easily achievable on consumer hardware, you can train something like Tacotron 2 + WaveRNN on your own computer to achieve similar, if not better
by enlyth 4y ago
This is already easily achievable on consumer hardware, you can train something like Tacotron 2 + WaveRNN on your own computer to achieve similar, if not better results. Check out this repo:
https://github.com/coqui-ai/TTS https://github.com/coqui-ai/TTS
You can also clone someone's voice by finetuning a pretrained LJSpeech model and training a vocoder from scratch, I've had great success with as little as 15 minutes of speech.
- davidzweig 4y agoCheck my other comment in this thread, you might try our dataset. :)
- enlyth 4y agoWill check it out, thanks
- GordonS 4y ago> You can also clone someone's voice by finetuning a pretrained LJSpeech model and training a vocoder from scratch, I've had great success with as little as 15 minutes of speech. Are you able to point to any articles to help get started with this please?
- enlyth 4y agoUnfortunately, I'm not aware of any beginner friendly tutorials. The way I learned it was just by experimenting with various GitHub repositories (e.g. https://github.com/fatchord/WaveRNN https://github.com/fatchord/WaveRNN or the one I linked earlier) but it takes a lot of trial and error. Might do a writeup at some point if I have time.
- rockemsockem 4y agoIt really isn't though. The level of fidelity that Apple is demonstrating in those samples is very impressive. You can generate fine voices with little work using repos like that one, but to get to the level Apple has takes a lot of work. EDIT: "fine voices" not "find voices"