All Models
K

OmniVoice

by k2-fsaPublicapache-2.0
README.md

OmniVoice

OmniVoice is a massively multilingual zero-shot text-to-speech, voice cloning, and voice design model family developed by k2-fsa on Hugging Face.

Links

  • Upstream model card: https://huggingface.co/k2-fsa/OmniVoice
  • Upstream GitHub: https://github.com/k2-fsa/sherpa-onnx

Model Facts

  • Task: Text-to-Speech, Zero-Shot Voice Cloning & Voice Design
  • Parameters: ~1B
  • License: Apache-2.0
  • Architecture: OmniVoice (GGUF)

LA Studio Notes

OmniVoice in LA Studio is a high-fidelity speech synthesis workflow. It downloads GGUF models from conversion repository Serveurperso/OmniVoice-GGUF. It needs:

  • Backbone: omnivoice-base-Q8_0.gguf
  • Tokenizer: omnivoice-tokenizer-F32.gguf

It supports over 600 languages, zero-shot voice cloning, and description-based voice design. It also supports non-verbal tags like [laughter], [sigh], etc.

Configure Workflow Parameters

Configure these custom parameters to fine-tune the speech and voice workflows.

Temperaturefloat

Controls randomness of speech generation. Higher values increase variety.

1
Seedint

Random seed for reproducibility. Set to -1 for random.

-1
Max Codec Stepsint

Maximum decoding steps for the neural voice audio synthesizer.

100