This is really quite good at sounding like Donald, especially for the first half of the audio. I’ll probably play around with this for a bit; it’s. It clear to me how much variation you can get in voice in latent space. Anyway it looks to be a very high quality (at least) short form tts engine with open weights so thanks team!
Is this really free software? I am really looking for _GOOD_ TTS software which is maintainable, really opensource (for every usage) and can do english/german/spanish/french/russian.
Zonos TTS is the SOTA, fully open-source (Apache license), and supports English, Japanese, Chinese, French, and German out of the box. You could train to add Russian, or run the output of this TTS through Meta's Seamless translation.
7 comments
[ 3.8 ms ] story [ 29.5 ms ] threadhttps://spark-tts.github.io/
https://github.com/Zyphra/Zonos
This is great as heavy TTS user. Waiting for real time 3x.