MiraTTS

This is the model for the MiraTTS repository. MiraTTS is a high quality TTS model that can generate clear and realistic speech at speeds as fast as 100x realtime.

Key benefits

  • Incredibly fast: Over 100x realtime by using Lmdeploy and batching.
  • High quality: Generates clear and crisp 48khz audio outputs which is much higher quality then most models.
  • Memory efficient: Works within 6gb vram.
  • Low latency: Latency can be low as 100ms.

Random samples, non cherry picked:

Thanks to Gapeleon for creating a great space for this model, you can try it here: https://huggingface.co/spaces/Gapeleon/Mira-TTS

If you find this model/code helpful, please give a like or star. Thank you.

Please check out the github repo for usage and finetuning notebooks.

Downloads last month
1,560
Safetensors
Model size
0.5B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for YatharthS/MiraTTS

Finetunes
1 model
Quantizations
1 model

Space using YatharthS/MiraTTS 1