SWivid / F5-TTS

Official code for "F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching"
https://arxiv.org/abs/2410.06885
MIT License
7.58k stars 946 forks source link

Training Tamil Language from Scratch #531

Open sam4muzix opened 2 days ago

sam4muzix commented 2 days ago

Checks

1. Is this request related to a challenge you're experiencing? Tell us your story.

I tried to finetune tamil dataset but results are partial. so i started to prepare for training from scratch. please help on how to train from scratch.

2. What is your suggested solution?

gradio webui for training from scratch.

3. Additional context or comments

No response

4. Can you help us with this feature?

ZhikangNiu commented 2 days ago
  1. prepare your dataset and your vocab
  2. accelerate config
  3. update your config in src/f5_tts/train/train.py
  4. accelerate launch src/f5_tts/train/train.py

check src/f5_tts/train/train.py and https://github.com/SWivid/F5-TTS/blob/d8e622f0043678c57c47ad520aed0265ee73f30b/src/f5_tts/train/train.py#L18-L37

i-am-flashh commented 1 day ago

hi @sam4muzix , have you got 1 speaker dataset for tamil? i guess multiple speaker wont work good.