This is an improvised video on how to train a new language for F5-TTS, currently the best open-source text-to-speech tool available.

Please subscribe to support my channel!

Original GitHub page: [ Ссылка ]
# F5 TTS French Model

This model was trained exclusively on French public domain material from [LibriVox]([ Ссылка ] and voice recordings. It is designed for text-to-speech synthesis in French and has gone through two rounds of training:

1. **Round 1**:
- **Samples**: 80k
- **Epochs**: 100
- **Speakers**: Single speaker

2. **Round 2**:
- **Samples**: 120k
- **Epochs**: 150
- **Speakers**: Multiple speakers from different corpus

### Model Checkpoint
The checkpoint supports French and can be [downloaded here] [ Ссылка ]

in the last version of F5tts in the web interface select "custom" and enter theses path:
MODEL_CKPT: hf://RASPIAUDIO/F5-French-MixedSpeakers-reduced/model_last_reduced.pt
VOCAB_FILE: hf://RASPIAUDIO/F5-French-MixedSpeakers-reduced/vocab.txt


### Online Inference
You can test the model online using our [Hugging Face Space] [ Ссылка ]

### Resources
- [Tutorial video to train a new language model] [ Ссылка ]
- [Discussion about this training can be found here] [ Ссылка ]

### About Us
Check out our website: [ Ссылка ] – We create cool, hackable audio devices that let you experiment with audio, voice assistants, and AI using ESP32 and Raspberry Pi!

свернуть