@herimor I managed to produce a lifelike realtime Hebrew version of the model - with only 3 hours of training and 160h of real speech data. Hebrew is morphologically complex and a rare language - yet the model adapted well, which proves its potential and versatility.
Feel free to reference/showcase the finetune in your work!
https://huggingface.co/flowty1/voxtream2-he
@herimor I managed to produce a lifelike realtime Hebrew version of the model - with only 3 hours of training and 160h of real speech data. Hebrew is morphologically complex and a rare language - yet the model adapted well, which proves its potential and versatility.
Feel free to reference/showcase the finetune in your work!
https://huggingface.co/flowty1/voxtream2-he