{
  "schemaVersion": "1.0",
  "updates": [],
  "companyNews": [],
  "id": "b891",
  "language": "zh",
  "name": "Kokoro",
  "originalName": "Kokoro",
  "library": "tools",
  "category": "音频音乐",
  "categoryKey": "音频音乐",
  "categoryKeys": [
    "音频音乐"
  ],
  "categories": [
    "音频音乐"
  ],
  "entryPoints": [],
  "summary": "82M 参数的开放权重语音合成模型，按语言和声音选择朗读文本，通过 Python 流程分段输出音频。",
  "introduction": "Kokoro 将待朗读文字、语言代码和预设声音传给 KPipeline，经过 Misaki 的字素到音素处理后生成音频，可逐段保存为 WAV。82M 权重适合在自建应用中集成，但仍需配置模型、声音文件和相应语言依赖；它不是任意上传参考音频即可复现音色的通用克隆界面。官方权重采用 Apache 2.0，语言与声音清单以对应版本为准，入口使用作者的模型卡。",
  "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/zh/products/kokoro--b891.html",
    "markdown": "read/zh/products/kokoro--b891.md",
    "json": "read/zh/products/kokoro--b891.json"
  }
}