tts / README.md
clementruhm's picture
Add info on uk models
0ebd3bd
|
raw
history blame
1.86 kB
metadata
language:
  - en
tags:
  - JETS
  - LightSpeech
  - MB-MelGAN
  - Text-to-Speech
datasets:
  - CMUArctic
  - Hi-Fi
pipeline_tag: text-to-speech

TTS Models

Here you can find models compatible with Balacoon software. There are several model types to be aware of:

  • *_jets_cpu.addon - JETS models for synthesis on CPU, compatible with balacoon_tts python package. A tutorial on how to use it can be found here. Those are high-end models, producing 24khz audio.
  • *_light_cpu.addon - Light models for lightning-fast synthesis on-device. Usage is the same as of JETS models, but the naturalness of synthesized audio is compromised in favor of speed. Models produce 16khz audio.
  • *_jets_gpu.addon - JETS models for synthesis on GPU, compatible with balacoon/tts_server docker image. A tutorial on how to use it can be found here. Exactly the same as jets_cpu models, but repacked for GPU.

You can check the interactive demo balacoon/tts space.

List of available models

  • en-US locale
  • uk locale
    • Ukrainian TTS datasets
      • uk_ltm_jets_cpu.addon
      • uk_ltm_jets_gpu.addon
      • uk_tetiana_light_cpu.addon - trained only on "Tetiana" speaker