Chanjing TTS Voice Clone#

技能包标识： chanjing-tts-voice-clone。

功能说明#

基于用户提供的 参考音频公网 URL（由蝉镜服务端拉取）创建音色并合成语音；轮询任务并从接口返回 URL 下载结果。脚本 不 依赖 ffmpeg/ffprobe。机器可读凭据声明见篇首 metadata.openclaw.credentialModel。 access_token 与 expire_in 会写入同一 credentials.json；刷新令牌时由本技能覆盖文件中对应字段，请勿将凭证提交到版本库。

运行依赖#

• python3 与同仓库 scripts/*.py
• 无 ffmpeg/ffprobe 门控

环境变量#

1. CHANJING_CONFIG_DIR（可选）： credentials.json 所在目录，默认 ~/.chanjing
2. CHANJING_API_BASE（可选）：Open API 基址，默认 https://open-api.chanjing.cc

使用命令#

• ClawHub（slug 以注册表为准）： clawhub run chanjing-tts-voice-clone
• 本仓库： python skills/chanjing-tts-voice-clone/scripts/create_voice.py …（见正文 How to Use）

---

登记摘要（英文 · ClawHub / OpenClaw）#

Primary credential (not primaryEnv): app_id + secret_key in ~/.chanjing/credentials.json (or $CHANJING_CONFIG_DIR/credentials.json); access_token and expire_in read/written in the same file ( overwritten on token refresh). primaryEnv omitted by design. metadata.openclaw.credentialModel is the machine-readable registry source of truth.

Required vs optional: CHANJING_API_BASE optional (default https://open-api.chanjing.cc). CHANJING_CONFIG_DIR optional. No ffmpeg/ ffprobe in skill metadata.

Purpose alignment: Voice clone TTS— user-supplied public URL for reference audio ( Chanjing servers fetch it); then synthesize and download from API-returned URLs. Trust API host, response URLs, and user-chosen reference URL.

See How to Use → Reviewer Q&A (four items) → Security & credentials (registry summary).

When to Use This Skill#

Use this skill when the user needs to generate speech from text, with a user-provided reference voice. The reference audio needs to be provided as a publicly accessible url.

This TTS service supports:

• bilingual Chinese and English
• adjustment of speech rate
• sentence-level timestamp

How to Use This Skill#

前置条件（权限验证）：执行本 Skill 前，必须先通过 chanjing-credentials-guard 完成 AK/SK 与 Token 校验。脚本与 guard 使用同一套凭证；无凭证时会执行 open_login_page.py 打开注册/登录页。

Reviewer Q&A (four items)#

Direct answers for Purpose & Capability / Instruction Scope / Credentials / Persistence-style reviews; aligned with code, top Registry summary (English), and description frontmatter.

#TopicAnswer1Purpose vs implementation; primary credentialAligned. Voice-clone TTS client; reference audio via user-provided public URL (server-side fetch). CHANJING_API_BASE not required. Primary credential in credentials.json; primaryEnv omitted. No ffmpeg/ ffprobe in metadata.2Runtime scope: secrets, URLsIn scope. Read/write credentials; browser if needed; HTTPS; user reference URL + output URLs from API—trust boundaries include user-chosen reference URL and API host.3Env vs file secretsCHANJING_* optional; app_id / secret_key / access_token / expire_in on disk—sensitive; token fields overwritten on refresh.4Persistence & privilegealways: false default; does not modify other skills or global agent config.

Security & credentials (registry summary)#

Aligned with Reviewer Q&A above, top English registry summary, and description.

AspectDetailsPrimary credentialSame file-based app_id / secret_key as other Chanjing skills. Not primaryEnv.TokenRefreshed and written to ~/.chanjing/credentials.json (or CHANJING_CONFIG_DIR); access_token and expire_in are overwritten when the skill refreshes the token.EnvOptional CHANJING_API_BASE, CHANJING_CONFIG_DIR.User-supplied URLCreate Voice API takes a public URL to reference audio; the user is responsible for that URL’s safety and legality; Chanjing’s servers retrieve it.Network / browserHTTPS to https://open-api.chanjing.cc; may open login when credentials are missing.DownloadsGenerated audio is downloaded from URLs returned by the API—trust the API host.PersistenceDefault always: false; does not modify other skills.

Chanjing-TTS-Voice-Clone provides an asynchronous speech synthesis API. Hostname for all APIs is: " https://open-api.chanjing.cc". All requests communicate using json. You should use utf-8 to encode and decode text throughout this task.

1. Obtain an access_token, which is required for subsequent requests
2. Call the Create Voice API, which accepts a url to an audio file as reference voice
3. Poll the Query Voice API until the result is success; keep the voice ID
4. Call the Create Speech Generation Task API, using the voice ID, record the task_id
5. Poll the Query Speech Generation Task Status API until the result is success
6. When the task status is complete, use the corresponding url in API response to download the generated audio file

Get Access Token API#

从 ~/.chanjing/credentials.json 读取 app_id 和 secret_key，若无有效 Token 则调用：

http

POST /open/v1/access_token
Content-Type: application/json

请求体（使用本地配置的 app_id、secret_key）：

json

{
  "app_id": "<从 credentials.json 读取>",
  "secret_key": "<从 credentials.json 读取>"
}

Response example:

json

{
  "trace_id": "8ff3fcd57b33566048ef28568c6cee96",
  "code": 0,
  "msg": "success",
  "data": {
    "access_token": "1208CuZcV1Vlzj8MxqbO0kd1Wcl4yxwoHl6pYIzvAGoP3DpwmCCa73zmgR5NCrNu",
    "expire_in": 1721289220
  }
}

Response field description:

First-level FieldSecond-level FieldDescriptioncodeResponse status codemsgResponse messagedataResponse dataaccess_tokenValid for one day, previous token will be invalidatedexpire_inToken expiration time

Response status code description:

codeDescription0Success400Parameter format error40000Parameter error50000System internal error

Create Voice API#

Post to the following endpoint to create a voice.

http

POST /open/v1/create_customised_audio
access_token: {{access_token}}
Content-Type: application/json

Request body example:

json

{
  "name": "example",
  "url": "https://example.com/abc.mp3"
}

Request field description:

FieldTypeRequiredDescriptionnamestringYesA name for this voiceurlstringYesurl to the reference audio file, format must be one of mp3, wav or m4a. Supported mime: audio/x-wav, audio/mpeg, audio/m4a, video/mp4. Size must not exceed 100MB. Recommended audio length: 30s-5minmodel_typestringYesUse "Cicada3.0-turbo"languagestringNoEither "cn" or "en", default to "cn"

Response example:

json

{
  "trace_id": "2f0f50951d0bae0a3be3569097305424",
  "code": 0,
  "msg": "success",
  "data": "C-Audio-53e4e53ba1bc40de91ffaa74f20470fc"
}

Response field description:

FieldDescriptioncodeStatus CodemsgMessagedataVoice ID, to be used in following steps

Response status code description:

CodeDescription0Success400Parameter Format Error10400AccessToken Error40000Parameter Error40001QPS Exceeded50000Internal System Error

Poll Voice API#

Send a GET request to the following endpoint to query whether the voice is ready to be used, voice ID is obtained in the previous step. The polling process may take a few minutes, keep polling until the status indicates the voice is ready.

http

GET /open/v1/customised_audio?id={{voice_id}}
access_token: {{access_token}}

Response example:

json

{
  "trace_id": "7994cedae0f068d1e9e4f4abdf99215b",
  "code": 0,
  "msg": "success",
  "data": {
    "id": "C-Audio-53e4e53ba1bc40de91ffaa74f20470fc",
    "name": "声音克隆",
    "type": "cicada1.0",
    "progress": 0,
    "audio_path": "",
    "err_msg": "不支持的音频格式，请阅读接口文档",
    "status": 2
  }
}

Response field description:

First-level FieldSecond-level FieldDescriptioncodeStatus CodemsgResponse MessagedataidVoice IDprogressProgress: range 0-100typenameerr_msgError Messageaudio_pathstatus0-queued; 1-in progress; 2-done; 3-expired; 4-failed; 99-deleted

Response status code description:

CodeDescription0Success10400AccessToken Error40000Parameter Error40001QPS Exceeded50000Internal System Error

Create Speech Generation Task API#

Post to the following endpoint to submit a speech generation task:

http

POST /open/v1/create_audio_task
access_token: {{access_token}}
Content-Type: application/json

Request body example:

json

{
  "audio_man": "C-Audio-53e4e53ba1bc40de91ffaa74f20470fc",
  "speed": 1,
  "pitch": 1,
  "text": {
    "text": "Hello, I am your AI assistant."
  }
}

Request field description:

Parameter NameTypeNested KeyRequiredExampleDescriptionaudio_manstringYesC-Audio-53e4e53ba1bc40de91ffaa74f20470fcVoice ID, obtained from previous stepspeednumberYes1Speech rate, range: 0.5 (slow) to 2 (fast)pitchnumberYes1Pitch (always set to 1)textobjecttextYesHello, I am your Cicada digital humanRich text, text length limit less than 4,000 charactersaigc_watermarkboolNofalseWhether to add visible watermark to audio, default is false

Response field description:

FieldDescriptioncodeResponse status codemsgResponse messagetask_idSpeech synthesis task ID

Example Response

json

{
  "trace_id": "dd09f123a25b43cf2119a2449daea6de",
  "code": 0,
  "msg": "success",
  "data": {
    "task_id": "88f635dd9b8e4a898abb9d4679e0edc8"
  }
}

Response status code description:

codeDescription0Success400Incoming parameter format error10400AccessToken verification failed40000Parameter error40001Exceeds QPS limit40002Production duration reached limit50000System internal error

Query Speech Generation Task Status API

Post a request to the following endpoint:

http

POST /open/v1/audio_task_state
access_token: {{access_token}}
Content-Type: application/json

Request body example:

json

{
  "task_id": "88f635dd9b8e4a898abb9d4679e0edc8"
}

Request field description:

Parameter NameTypeRequiredExampleDescriptiontask_idstringYes88f789dd9b8e4a121abb9d4679e0edc8Task ID obtained in previous step

Response body example:

json

{
  "trace_id": "ab18b14574bbcc31df864099d474080e",
  "code": 0,
  "msg": "success",
  "data": {
    "id": "9546a0fb1f0a4ae3b5c7489b77e4a94d",
    "type": "tts",
    "status": 9,
    "text": [
      "猫在跌落时能够在空中调整身体，通常能够四脚着地，这种”猫右自己“反射显示了它们惊人的身体协调能力和灵活性。核磁共振成像技术通过利用人体细胞中氢原子的磁性来生成详细的内部图像，为医学诊断提供了重要工具。"
    ],
    "full": {
      "url": "https://cy-cds-test-innovation.cds8.cn/chanjing/res/upload/tts/2025-04-08/093a59021d85a72d28a491f21820ece4.wav",
      "path": "093a59013d85a72d28a491f21820ece4.wav",
      "duration": 18.81
    },
    "slice": null,
    "errMsg": "",
    "errReason": "",
    "subtitles": [
      {
        "key": "20c53ff8cce9831a8d9c347263a400a54d72be15",
        "start_time": 0,
        "end_time": 2.77,
        "subtitle": "猫在跌落时能够在空中调整身体"
      },
      {
        "key": "e19f481b6cd2219225fa4ff67836448e054b2271",
        "start_time": 2.77,
        "end_time": 4.49,
        "subtitle": "通常能够四脚着地"
      },
      {
        "key": "140beae4046bd7a99fbe4706295c19aedfeeb843",
        "start_time": 4.49,
        "end_time": 5.73,
        "subtitle": "这种，猫右自己"
      },
      {
        "key": "e851881271876ab5a90f4be754fde2dc6b5498fd",
        "start_time": 5.73,
        "end_time": 7.97,
        "subtitle": "反射显示了它们惊人的身体"
      },
      {
        "key": "fbb0b4138bad189b9fc02669fe1f95116e9991b4",
        "start_time": 7.97,
        "end_time": 9.45,
        "subtitle": "协调能力和灵活性"
      },
      {
        "key": "f73404d135feaf84dd8fbea13af32eac847ac26d",
        "start_time": 9.45,
        "end_time": 12.49,
        "subtitle": "核磁共振成像技术通过利用人体"
      },
      {
        "key": "e18827931223962e477b14b2b8046947039ac222",
        "start_time": 12.49,
        "end_time": 14.77,
        "subtitle": "细胞中氢原子的磁性来生成"
      },
      {
        "key": "d137bf2b0c8b7a39e3f6753b7cf5d92bd877d2d9",
        "start_time": 14.77,
        "end_time": 15.97,
        "subtitle": "详细的内部图像"
      },
      {
        "key": "0773911ae0dbaa763a64352abdb6bdac3ff8f149",
        "start_time": 15.97,
        "end_time": 18.41,
        "subtitle": "为医学诊断提供了重要工具"
      }
    ]
  }
}

Response field description:

First-level FieldSecond-level FieldThird-level FieldDescriptioncodeResponse status codemsgResponse messagedataidAudio IDtypeSpeech typestatusStatus: 1 - in progress, 9 - donetextSpeech textfullurlurl to download generated audio filepathdurationAudio durationsliceerrMsgError messageerrReasonError reasonsubtitles(array type)keySubtitle IDstart_timeSubtitle start time pointend_timeSubtitle end time pointsubtitleSubtitle text

Response field description:

CodeDescription0Response successful10400AccessToken verification failed40000Parameter error50000System internal error

Scripts#

本 Skill 提供脚本（ skills/chanjing-tts-voice-clone/scripts/），与 chanjing-credentials-guard 使用同一配置文件；无 AK/SK 时会 执行 open_login_page.py 脚本 打开注册/登录页。

脚本说明create_voice.py提交定制声音任务（参考音频 URL），输出 voice_idpoll_voice.py轮询定制声音直到就绪（status=2），输出 voice_idcreate_task.py使用定制声音创建 TTS 任务，输出 task_idpoll_task.py轮询 TTS 任务直到完成，输出音频下载 URL

示例（在项目根或 skill 目录下执行）：

bash

# 1. 创建定制声音（参考音频需为公开 URL）
VOICE_ID=$(python skills/chanjing-tts-voice-clone/scripts/create_voice.py --name "我的声音" --url "https://example.com/ref.mp3")

# 2. 轮询直到声音就绪
python skills/chanjing-tts-voice-clone/scripts/poll_voice.py --voice-id "$VOICE_ID"

# 3. 创建 TTS 任务
TASK_ID=$(python skills/chanjing-tts-voice-clone/scripts/create_task.py --audio-man "$VOICE_ID" --text "Hello, I am your AI assistant.")

# 4. 轮询到完成，得到音频下载链接
python skills/chanjing-tts-voice-clone/scripts/poll_task.py --task-id "$TASK_ID"