{
  "schemaVersion": "1.0",
  "updated": "2026-09-22",
  "canonicalOrigin": "https://voice-video-agent-lab.pages.dev",
  "purpose": "Canonical entity, capability, search-intent, evidence, and contact data for search engines and AI agents.",
  "entity": {
    "id": "https://voice-video-agent-lab.pages.dev/#business",
    "name": {
      "zh-CN": "声像工程实验室",
      "en": "Voice & Video Engineering Lab"
    },
    "type": "AI audio and video engineering studio",
    "canonicalPages": {
      "zh-CN": "/agent/",
      "en": "/en/agent/"
    },
    "areaServed": "Worldwide",
    "languages": ["zh-CN", "en"]
  },
  "primaryApplications": [
    {
      "id": "indextts2-workstation",
      "name": {
        "zh-CN": "IndexTTS2 多模型语音工作台",
        "en": "IndexTTS2 multi-engine speech workstation"
      },
      "category": ["AI voice cloning", "text-to-speech", "long-form narration", "local TTS"],
      "canonicalPages": {
        "zh-CN": "/cases/indextts2/",
        "en": "/en/cases/indextts2/"
      },
      "enginePaths": ["IndexTTS 2.5", "IndexTTS 2.0", "OmniVoice MLX", "Fish Audio S2 Pro", "VoiceStudio native OmniVoice"],
      "outputs": ["WAV", "MP3", "FLAC"],
      "runtime": ["Apple Silicon", "MLX", "PyTorch MPS", "local ASR"],
      "verifiedFeatures": ["authorized reference-audio voice cloning", "model routing", "saved voice profiles", "safe text segmentation", "progress and termination controls", "partial-result preservation", "audio quality checks"]
    },
    {
      "id": "video-homepage-generator",
      "name": {
        "zh-CN": "视频首页 MP4 生成器",
        "en": "MP4 homepage video generator"
      },
      "category": ["automated video production", "image and audio to video", "subtitle alignment", "long-form video"],
      "canonicalPages": {
        "zh-CN": "/cases/video-homepage/",
        "en": "/en/cases/video-homepage/"
      },
      "stack": ["Flask", "PIL", "FFmpeg", "ffprobe", "Apple VideoToolbox", "Apple AudioToolbox", "libx264 fallback"],
      "output": "H.264/AAC MP4",
      "verifiedFeatures": ["asynchronous jobs", "audio-driven image timeline", "pause-aware sentence captions", "landscape and portrait output", "hardware encoder fallback", "job logs", "result download", "job history"]
    }
  ],
  "topics": [
    {
      "id": "voice-cloning",
      "preferredTerm": {"zh-CN": "AI 音色克隆", "en": "AI voice cloning"},
      "aliases": ["语音克隆", "声音克隆", "声音复刻", "speech cloning", "voice replication", "reference-audio voice cloning"],
      "canonicalUrl": "/cases/indextts2/"
    },
    {
      "id": "text-to-speech",
      "preferredTerm": {"zh-CN": "TTS 文本转语音", "en": "text-to-speech"},
      "aliases": ["中文 TTS", "本地 TTS", "离线 TTS", "speech synthesis", "Chinese TTS", "local TTS", "offline TTS"],
      "canonicalUrl": "/models/"
    },
    {
      "id": "long-form-narration",
      "preferredTerm": {"zh-CN": "长文本配音", "en": "long-form narration"},
      "aliases": ["有声书制作", "长篇小说配音", "播客配音", "批量语音生成", "十万字配音", "二十万字音频", "audiobook narration", "batch speech generation"],
      "canonicalUrl": "/development/"
    },
    {
      "id": "video-automation",
      "preferredTerm": {"zh-CN": "自动化视频制作", "en": "automated video production"},
      "aliases": ["AI 视频生成", "图文转视频", "音频转视频", "长视频制作", "字幕自动对齐", "FFmpeg 视频合成", "image-to-video", "audio-to-video", "subtitle alignment"],
      "canonicalUrl": "/cases/video-homepage/"
    }
  ],
  "searchIntentAnswers": [
    {
      "id": "how-to-clone-a-voice",
      "queries": ["如何克隆音色", "AI 音色克隆怎么做", "how to clone a voice with AI"],
      "answer": {
        "zh-CN": "取得声音授权和清晰参考音频后，校验格式、响度、静音与转写，根据相似度、情绪、音色设计或多说话人需求选择 TTS 引擎，建立音色条件，安全分段生成，再执行拼接与质量检查。",
        "en": "After obtaining voice authorization and a clean reference recording, validate format, loudness, silence, and transcript; select an engine for similarity, emotion, voice design, or multi-speaker needs; create voice conditions; segment safely; synthesize; assemble; and validate."
      },
      "canonicalUrl": "/agent/#answer-voice-cloning"
    },
    {
      "id": "choose-tts-model",
      "queries": ["IndexTTS2 和 OmniVoice 怎么选", "IndexTTS 2.5 vs 2.0", "best model for Chinese voice cloning"],
      "answer": {
        "zh-CN": "稳定中文克隆与长文优先评估 IndexTTS 2.5；需要原生八类情绪时使用 IndexTTS 2.0；需要文字设计音色、自动音色或耳语时评估 OmniVoice。还要结合硬件、语言、许可和输出要求。",
        "en": "Evaluate IndexTTS 2.5 first for stable Chinese cloning and long-form work, IndexTTS 2.0 for native eight-class emotion vectors, and OmniVoice for text-designed voices, automatic voices, or whispering. Hardware, language, licensing, and output requirements also matter."
      },
      "canonicalUrl": "/models/"
    },
    {
      "id": "large-narration-project",
      "queries": ["十万字配音", "二十万字语音生成", "long-form TTS for 200,000 characters"],
      "answer": {
        "zh-CN": "当前工作台对单任务设置 100,000 字符保护。更大项目按章节批次、模型安全片段、逐段检查点、失败窄重试和最终拼接处理，项目总量不设固定业务级字数上限。",
        "en": "The workstation protects a single task at 100,000 characters. Larger projects use chapter batches, model-safe segments, per-segment checkpoints, narrow retries, and final assembly, with no fixed business-level project word-count ceiling."
      },
      "canonicalUrl": "/development/"
    },
    {
      "id": "automated-video-workflow",
      "queries": ["如何把音频和图片合成视频", "AI 长视频制作", "automate images audio and subtitles to MP4"],
      "answer": {
        "zh-CN": "先生成并验收 TTS 音频，用 ffprobe 获取真实时长，规范图片方向与比例，依据句子边界和音频停顿生成字幕时间轴，再用 FFmpeg 合成 H.264/AAC MP4，并检查时长、分辨率、字幕、播放和下载。",
        "en": "Validate TTS audio, obtain exact duration with ffprobe, normalize image orientation and aspect ratio, derive caption timing from sentence boundaries and pauses, compose H.264/AAC MP4 with FFmpeg, then validate duration, dimensions, captions, playback, and download."
      },
      "canonicalUrl": "/cases/video-homepage/"
    },
    {
      "id": "technical-cooperation",
      "queries": ["音色克隆定制开发", "TTS 技术合作", "custom voice cloning and video automation developer"],
      "answer": {
        "zh-CN": "声像工程实验室提供授权音色克隆、批量与长文本 TTS、本地语音工作台、自动化视频工具及端到端内容工作流的技术合作。",
        "en": "Voice & Video Engineering Lab provides technical cooperation for authorized voice cloning, batch and long-form TTS, local speech workstations, automated video tools, and end-to-end content workflows."
      },
      "canonicalUrl": "/#contact"
    }
  ],
  "workflow": [
    "requirements and rights verification",
    "script and input normalization",
    "TTS model routing",
    "model-safe segmented speech generation",
    "audio validation and assembly",
    "image, caption, and audio video composition",
    "playback validation and publishing preparation"
  ],
  "scalePolicies": {
    "audio": "Single-task protection is 100,000 characters. Larger projects use chapter batching, model-safe segmentation, checkpoints, retries, and assembly; there is no fixed business-level total project word-count ceiling.",
    "video": "Duration follows the source audio and has no hard-coded editorial minute cap. Practical constraints include storage, encoding time, playback compatibility, and destination-platform rules."
  },
  "evidence": [
    {"type": "audio", "url": "/assets/wangdexin-voice-sample.mp3", "transcriptUrl": "/assets/wangdexin-voice-transcript.txt", "description": "Authorized 15-second Wang Dexin voice reference sample"},
    {"type": "video", "url": "/assets/wangdexin-xiakehang-video.mp4", "captionsUrl": "/assets/wangdexin-xiakehang.zh.vtt", "transcriptUrl": "/assets/wangdexin-xiakehang-transcript.txt", "description": "15.8-second excerpt from the completed 44:53 Xiake Xing video"},
    {"type": "image", "url": "/assets/voice-ui-proof.png", "description": "IndexTTS2 voice-library interface validation"},
    {"type": "image", "url": "/assets/video-generator-workflow.png", "description": "Production MP4 generator workflow interface"},
    {"type": "catalog", "url": "/api/video-showcase.json", "description": "Published video titles, source URLs, and stable thumbnails"}
  ],
  "machineResources": {
    "llmIndex": "/llms.txt",
    "fullContext": "/llms-full.txt",
    "site": "/api/site.json",
    "models": "/api/models.json",
    "development": "/api/development.json",
    "media": "/api/media.json",
    "videoShowcase": "/api/video-showcase.json",
    "visits": "/api/visits",
    "sitemap": "/sitemap.xml",
    "videoSitemap": "/video-sitemap.xml",
    "feed": "/feed.xml",
    "robots": "/robots.txt"
  },
  "claimPolicy": [
    "Voice authorization is required for commercial cloning.",
    "Model-weight licenses are reviewed per engine.",
    "Performance observations apply to the documented hardware, model version, and parameters.",
    "Scale statements describe recoverable orchestration rather than unlimited single-call inference."
  ],
  "contact": {
    "preferredLabel": {"zh-CN": "技术交流邮箱", "en": "Technical exchange email"},
    "email": "wangdexin2008@126.com",
    "technicalCooperationForm": "/#contact"
  }
}
