Will there be more languages?
Sorry, this is a bit selfish post.
I love the product.
I can't speak for other languages because I only know my native language and English. I am a Bengali speaker. Still, there is no TTS model that has a well-voiced Bengali voice. I have used several TTS models like OpenAI's advanced, Gemini, and Azure, but they still sound robotic, not as natural as when spoken in English.
Fun fact: If you search the top 10 most popular languages spoken around the world, Bengali is one of them.
We WANT to add more language, but there's some problems we are facing:
- We need enough data for pretraining
- We need more users to use the language
...
So this will be a long time job.
PoTaTo721, I tested the S1 (not a mini) cloning on the fish audio website for Russian. There is a problem with the accent in the recordings, the swallowing of some letters, as well as other artifacts. Where can I send examples of transcribed text recordings to fix problems? I also wonder when it will be possible to train my own model on S1? Are you planning to release the full version of the S1 model in the open source? Side question: Is it possible to use the S1 mini in the f5 tts, or are the architectures incompatible?