Skip to content

Kobold Support

erew123 edited this page Sep 30, 2024 · 1 revision

KoboldCPP offers native built-in support for AllTalk, providing an integrated text-to-speech solution. There are some limitations and considerations to be aware of when using this integration.

Limitations

  1. Streaming Generation Only:

    • The current integration only supports streaming generation.
    • This limits the use to Coqui XTTS models exclusively.
  2. Browser Compatibility:

    • Firefox is not compatible with the PCM stream generated by the XTTS model.
    • Firefox users will encounter an error: NS_ERROR_DOM_MEDIA_METADATA_ERR (0x806e0006) in their Developer console.
    • It is recommended to use alternative browsers such as Chrome, Edge, or Safari for the best experience.

Setup and Usage

  1. Ensure you have KoboldCPP installed and configured properly.
  2. Set up AllTalk with the Coqui XTTS model.
  3. In the KoboldCPP interface, locate the AllTalk integration settings.
  4. Enable AllTalk and configure any necessary settings (e.g., voice selection, language).
  5. Start your text generation session in KoboldCPP.
  6. The generated text should now be automatically converted to speech in real-time.

Troubleshooting

  • If you're not hearing any audio, ensure your system's audio output is working correctly.
  • Check that you're using a compatible browser (avoid Firefox due to the known issue).
  • Verify that the AllTalk service is running and accessible to KoboldCPP.
  • If issues persist, check the KoboldCPP console for any error messages related to AllTalk integration.
Clone this wiki locally