@bjelser/tts-compatible

OpenAI-compatible TTS provider for @bjelser/kit. Drives any TTS endpoint that speaks the OpenAI /v1/audio/speech shape — self-hosted Kokoro servers, OpenWebUI endpoints, or any community-built OpenAI-compatible bridge.

This is the lever for callers who need PyTorch Kokoro parity or any other server-side TTS not directly wrapped by a dedicated @bjelser/tts-* package.

Install

npm install @bjelser/tts-compatible
# or
bun add @bjelser/tts-compatible

No peer dependency — this provider talks to its endpoint via plain fetch.

Use

// docent.config.ts
import {corePlugins} from '@bjelser/core';
import {compatibleTtsProvider} from '@bjelser/tts-compatible';

export default {
  plugins: [...corePlugins, compatibleTtsProvider],
};

// films/my-film.json
{
  "meta": {
    "voice": {
      "provider": "openai-compatible",
      "endpoint": "http://localhost:8000/v1/audio/speech",
      "voice": "af_heart",
      "model": "kokoro"
    }
  }
}

If the endpoint expects an Authorization: Bearer … header, set OPENAI_COMPATIBLE_API_KEY.

Capabilities

• nativeAlignment: null (chunk-level only — OpenAI-shaped endpoints don't expose word timings)
• voiceCloning: depends on the underlying server
• languages: depends on the underlying server

License

MIT