Once you have the API running, the base url will depend on whether you are running SillyTavern itself from Docker, or directly (if you used their installer, it's likely running directly).

1. Go to the Extensions menu in the top bar of SillyTavern.

   

2. Find the TTS menu on that page and where it asks you to Select TTS Provider, select OpenAI Compatible

   

3. Click the Enabled box, and if you want it running automatically, Narrate user messages and/or Auto Generation, etc.

   

4. Fill in the provider endpoints:

   • If running directly: http://localhost:8880/v1/audio/speech
   • If running in Docker: http://host.docker.internal:8880/v1/audio/speech

5. You may have to hit the Reload button (the one right next to where it says OpenAI Compatible) to have it open up the remaining settings.

   

6. Set the model as kokoro, the API Key as not-needed

7. In available voices: You can either use the known listed voices of the Kokoro model by name, or there is an auto-mapping of voices in the OpenAI (SillyTavern prefills these in as well). For auto-narration, you have to manually edit this line to map which character gets which voice:

   Remsky:alloy,Assistant:echo,fable,onyx,nova,shimmer
   

   

Voila! You should now be able to go back to chat and select the option you want (e.g., "Narrate All Chat"), and it should start speaking.