← Back to Skills Marketplace
audiobooklm

audiobooklm-podcast

by audiobooklm · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
35
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install audiobooklm-podcast
Description
基于新音剪创作平台以及珠峰AI自研音色,提供播客创作、内容管理、音频生成、专辑上架等功能。适用于用户提出"播客生成"、"播客制作"等需求。
README (SKILL.md)

audiobooklm-podcast

audiobooklm_mcp 的标准播客/有声书生产链路。用户提供文本(或让用户先生成),最终产出可上架的章节音频。


核心原则

  1. 每一步可选择的操作都要征求用户意见
  2. 工具调用前必须先问用户确认,禁止自动决定
  3. 使用播客类音色时优先选择 category 为"播客"的音色
  4. 整章合成用 chapter_task_create,不要逐段 task_create
  5. 先混音再上架,顺序不能乱

Step 0 · 环境自检

调用任意 mcp__audiobooklm_mcp__* 工具时,如果返回 401/403/unauthorized/token expired/connection refused 等错误,暂停业务流程并提示用户配置 MCP。

配置指引:

看起来你还没接入 audiobooklm_mcp,按以下 4 步完成配置:

  1. 注册/登录 https://aigc.ximalaya.com/user
  2. 创建制作组,生成 API Token
  3. 把 MCP 服务写入配置:
{
  "mcpServers": {
    "audiobooklm_mcp": {
      "type": "http",
      "url": "https://aigc.ximalaya.com/audiobooklm/mcp",
      "headers": {
        "Authorization": "Bearer \x3CToken>"
      }
    }
  }
}
  1. 重启服务,确认mcp已连接成功(需要结合当前环境给出具体的重启提示)

标准流程

1. 确认信息

开始前必须确认以下信息:

字段 是否必需 处理方式
文本内容 必需 用户已给直接用;让你生成则先出文案确认
书籍 必需 询问:新建还是用已有?
章节名 必需 询问用户确认
文本形态 必需 普通旁白/单人播客 vs 多人对话(决定导入工具)
音色 必需 调用 list_tts_voices 向用户展示候选
专辑 可选 仅上架时需要,询问:新建/已有/不上架

2. 确定书籍

询问用户:新建书籍还是用已有书籍?

  • 新建 → 向用户确认书名后调用:

    book_create(book_name="\x3C书名>", book_type="podcast")
    
  • 已有 → 用 read_abs(scope={"domain":"books"}) 拉取书架,列出最近 10 本让用户选择


3. 创建章节

根据文本形态选择工具:

普通旁白/单人播客 → podcast_import

podcast_import(
  book_id=\x3Cbook_id>,
  chapter_name="\x3C章节名>",  # 新建章节时必填
  script_lines=["文本1", "文本2", ...], # 单人播客只需要传入文本列表,不需要角色标识
)

多人对话 → podcast_import

podcast_import(
  book_id=\x3Cbook_id>,
  chapter_name="\x3C章节名>",  # 新建章节时必填
  script_lines=["[角色名] 文本1", "[角色名] 文本2", ...] # 多人播客需要给出每段文本的角色名
)

注意:

  • 多人播客每行格式必须为 [角色名] 文本
  • 单人播客每行只需要文本,不需要角色名,格式为文本
  • chapter_name 和 chapter_id 二选一:新建章节传 chapter_name,追加到已有章节传 chapter_id
  • 返回 chapter_id 和 characters (角色名与 character_id 映射)

4. 列出音色

调用 list_tts_voices() 获取音色列表。

筛选建议:

  • 播客类优先选 category 包含"播客"的音色
  • 根据角色特征推荐合适的音色给用户选择
  • 向用户展示 3-5 个候选 + 简短描述

询问用户:每个角色分别选哪个音色?


5. 绑定音色到角色

mcp__audiobooklm_mcp__patch_abs(
  scope={"domain": "book", "book_id": \x3Cbook_id>},
  operations=[
    {
      "op_id": "op-\x3C角色序号>",
      "type": "update_character",
      "character_id": \x3Ccharacter_id>,
      "patch": {"speaker_id": \x3Cspeaker_id>},
      "reason": "绑定音色"
    }
    // 多个角色用逗号分隔
  ]
)

⚠️ 注意

  • 每个 operation 都要有唯一 op_id
  • type 是 "update_character" 不是 "op"

6. 整章合成

使用 chapter_task_create 一次提交整章:

mcp__audiobooklm_mcp__chapter_task_create(
  book_id=\x3Cbook_id>,
  chapter_id=\x3Cchapter_id>
)

返回 task_ids 列表。


7. 查询任务结果

mcp__audiobooklm_mcp__task_query(
  task_ids=[\x3Ctask_ids列表>],
  wait_timeout=120  # 可根据段落数调整
)

所有段落 status=succeeded 后继续。


8. 章节混音

mcp__audiobooklm_mcp__chapter_audio_mix(
  book_id=\x3Cbook_id>,
  chapter_id=\x3Cchapter_id>
)

⚠️ 顺序要求:必须先完成第 7 步,再调用混音。

返回 audio_url 即章节完整音频。


9. 上架专辑(可选)

询问用户:需要上架到专辑吗?新建/已有/不上架?

  • 新建专辑 → 询问专辑名后调用:

    mcp__audiobooklm_mcp__create_album(
      album_name="\x3C专辑名>",
      book_content="\x3C简介>"  # 可选,让 AI 生成封面
    )
    
  • 已有专辑 → 用户提供 album_id

  • 上架

    mcp__audiobooklm_mcp__upload_audio_to_album(
      album_id=\x3Calbum_id>,
      audio_url=\x3C第8步返回的audio_url>
    )
    

交付内容

完成后告诉用户:

  • 书籍 ID / 章节 ID
  • 选用的音色 (speakerId + 名称)
  • 章节完整音频 URL
  • 编辑页面:https://aigc.ximalaya.com/edit/\x3Cbook_id>/\x3Cchapter_id>
  • 账单或充值页面: https://aigc.ximalaya.com/payment

常见问题

现象 原因 解决
找不到 mcp__audiobooklm_mcp__* MCP 未注册 配置 MCP
401/403/unauthorized Token 无效/过期 重新生成 Token
patch_abs validation_failed 缺 op_id 或 type 检查 operation 格式
task_query 返回 processing 异步任务未完成 继续轮询 task_query
音色不对 character 未绑定 speaker 重新绑定步骤
create_album 失败 需要 book_content 传入简介文本

多人对话模板

适用于用户已提供:book_id + 章节名 + 对话脚本(格式为 [角色名] 文本

流程

  1. podcast_import 导入章节 → 获得 chapter_id 和 characters
  2. list_tts_voices 列出音色 → 用户为每个角色选择
  3. patch_abs 绑定音色 → 每个角色都要绑
  4. chapter_task_create 提交整章
  5. task_query 查询结果
  6. chapter_audio_mix 混音
Usage Guidance
Before installing, make sure you trust the Ximalaya/AIGC MCP endpoint and understand that the skill can create, synthesize, and optionally upload podcast content in your account. Keep the API token private, approve each tool call only after checking the target book/chapter/album, and avoid submitting confidential scripts unless you are comfortable with the provider processing them.
Capability Analysis
Type: OpenClaw Skill Name: audiobooklm-podcast Version: 1.0.0 The skill provides a standard workflow for generating podcasts and audiobooks using the Ximalaya AIGC platform (aigc.ximalaya.com). It utilizes MCP tools for book management, voice synthesis, and audio mixing, while explicitly requiring user confirmation for all tool executions. No indicators of malicious intent, data exfiltration, or unauthorized execution were found in SKILL.md or the metadata.
Capability Tags
cryptocan-make-purchases
Capability Assessment
Purpose & Capability
The described capabilities—podcast creation, content management, audio generation, and album publishing—match the stated purpose. They are still account-mutating actions that may involve provider billing or credits, so users should review confirmations carefully.
Instruction Scope
The skill gives a step-by-step workflow and explicitly requires user confirmation before tool calls and before optional choices such as creating books, choosing voices, or publishing to an album.
Install Mechanism
There is no install spec or code, but the SKILL.md instructs users to configure an external MCP endpoint with a Bearer token. The registry metadata does not declare a primary credential, so users should notice this setup requirement.
Credentials
Use of the Ximalaya/AIGC remote MCP service is proportionate to the skill’s purpose, but user text, generated audio, and account authorization will be handled by that external service.
Persistence & Privilege
The workflow creates persistent provider-side objects such as books, chapters, voice bindings, audio URLs, and optional album uploads. The artifacts present these as user-directed actions rather than hidden background behavior.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install audiobooklm-podcast
  3. After installation, invoke the skill by name or use /audiobooklm-podcast
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of audiobooklm-podcast. - Provides podcast creation, content management, audio generation, and album publishing based on the 新音剪 platform and custom voices. - Guides users through every step, requiring user confirmation for all actions. - Supports both single-voice narration and multi-person podcast formats. - Ensures correct workflow order: text entry, book selection, chapter import, voice selection, synthesis, mixing, and optional album publishing. - Includes detailed user instructions for configuration and troubleshooting common issues.
Metadata
Slug audiobooklm-podcast
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is audiobooklm-podcast?

基于新音剪创作平台以及珠峰AI自研音色,提供播客创作、内容管理、音频生成、专辑上架等功能。适用于用户提出"播客生成"、"播客制作"等需求。 It is an AI Agent Skill for Claude Code / OpenClaw, with 35 downloads so far.

How do I install audiobooklm-podcast?

Run "/install audiobooklm-podcast" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is audiobooklm-podcast free?

Yes, audiobooklm-podcast is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does audiobooklm-podcast support?

audiobooklm-podcast is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created audiobooklm-podcast?

It is built and maintained by audiobooklm (@audiobooklm); the current version is v1.0.0.

💬 Comments