{
  "schemaVersion": "1.0",
  "updates": [],
  "companyNews": [],
  "id": "b891",
  "language": "en",
  "name": "Kokoro",
  "originalName": "Kokoro",
  "library": "tools",
  "category": "Audio & music",
  "categoryKey": "音频音乐",
  "categoryKeys": [
    "音频音乐"
  ],
  "categories": [
    "Audio & music"
  ],
  "entryPoints": [],
  "summary": "An 82M-parameter open-weight TTS model that reads text with language and voice selections and produces audio chunks through Python.",
  "introduction": "Kokoro’s KPipeline accepts text, a language code and a preset voice. Misaki handles grapheme-to-phoneme processing, and generated audio chunks can be saved as WAV. Integration requires model and voice files plus language dependencies; this workflow selects provided voices rather than cloning arbitrary reference audio. Official weights use Apache 2.0. Language and voice inventories depend on the version, and the entry opens the author’s model card.",
  "capabilities": [],
  "company": {
    "name": "hexgrad",
    "originalName": "hexgrad"
  },
  "officialUrl": "https://huggingface.co/hexgrad/Kokoro-82M",
  "profileUpdatedOn": "2026-10-22",
  "sources": [
    {
      "kind": "product",
      "url": "https://huggingface.co/hexgrad/Kokoro-82M"
    },
    {
      "kind": "documentation",
      "url": "https://github.com/hexgrad/kokoro"
    }
  ],
  "news": [],
  "links": {
    "interactive": "products/kokoro--b891.html",
    "reading": "read/en/products/kokoro--b891.html",
    "markdown": "read/en/products/kokoro--b891.md",
    "json": "read/en/products/kokoro--b891.json"
  }
}