← 返回 Skills 市场
jackyken

CreateVideo - Podcast to Video

作者 Conan · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
86
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install create-podcast-video
功能描述
视频生成工具。当用户说"CreateVideo"、"创建视频"、"生成视频"或提供文案要求制作播客视频时触发。支持双人播客音频生成(通过 ListenHub MCP)、模版视频裁剪合并、内容分析输出。依赖 ffmpeg 和 ListenHub MCP Server。
使用说明 (SKILL.md)

CreateVideo Skill

工作流程

1. 接收输入

用户提供:

  • 必须:视频文案内容
  • 可选:模版选择(默认根据文案内容自动选择)
    • 书房 / 夜景$BASE_DIR/v_template.mp4(上海夜景书房,29.6s)
    • 能量 / 能量人$BASE_DIR/v_template_energy.mov(能量人,14.7s)

2. 创建任务目录

DATE=$(date +%Y%m%d)
BASE_DIR="/root/.openclaw/workspace/video-projects"
TASK_DIR="$BASE_DIR/${DATE}_001"

COUNTER=1
while [ -d "$TASK_DIR" ]; do
  COUNTER=$((COUNTER + 1))
  TASK_DIR="$BASE_DIR/${DATE}_$(printf "%03d" $COUNTER)"
done

mkdir -p "$TASK_DIR/temp"

3. 复制模板视频

根据文案内容或用户指定选择模版:

# 自动选择逻辑:能量/潜意识/吸引力类 → 能量人模版,其他 → 书房夜景模版
# 用户也可以直接指定:"用能量人模版" / "用书房模版"
cp "$BASE_DIR/v_template.mp4" "$TASK_DIR/template.mp4"        # 书房夜景(默认)
# 或
cp "$BASE_DIR/v_template_energy.mov" "$TASK_DIR/template.mp4"  # 能量人

4. 生成双人播客音频(ListenHub MCP)

通过 ListenHub MCP 生成双人播客音频:

  1. 调用 create_podcast
    • query: 用户文案内容
    • speakerIds: ["YOUR_SPEAKER_ID_1", "YOUR_SPEAKER_ID_2"](主播A先说,主播B后说)
    • language: "zh"
    • mode: "quick"
  2. 等待生成完成,下载音频到 $TASK_DIR/podcast_audio.mp3

详细 MCP 参数参见 references/listenhub-guide.md

5. 无缝循环模版视频(stream_loop)

使用 ffmpeg 的 stream_loop 实现无缝循环,一步完成循环+音频合并+输出:

⚠️ 不要用 concat demuxer + 正放↔倒放方案,会在接缝处冻帧(2026-03-22 踩坑验证)

DURATION=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$TASK_DIR/podcast_audio.mp3")
TEMPLATE_DUR=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$TASK_DIR/template.mp4")
LOOPS=$(python3 -c "import math; print(math.ceil($DURATION / $TEMPLATE_DUR))")

TITLE="根据文案内容生成的标题"

ffmpeg -y -stream_loop $LOOPS -i "$TASK_DIR/template.mp4" \
  -i "$TASK_DIR/podcast_audio.mp3" \
  -t $DURATION \
  -c:v libx264 -preset fast -b:v 1400k \
  -c:a aac -map 0:v:0 -map 1:a:0 \
  "$TASK_DIR/${TITLE}.mp4"

说明:

  • stream_loop 直接循环原始模版,无接缝问题
  • -b:v 1400k 控制输出大小(720×1280 竖屏约 40MB/4min)
  • 一条命令同时完成:循环视频 + 合并音频 + 裁剪时长 + 输出最终文件
  • 不再需要中间文件(trimmed.mp4、merged.mp4、template_reversed.mp4)

6. 生成内容分析

基于原始文案和播客脚本,生成并保存到 $TASK_DIR/内容分析.txt

  1. 核心金句 — 最具传播力的一句话
  2. 洞见 — 深层洞察和价值点
  3. 内容描述 — 50字概括视频核心内容
  4. 朋友圈文案 — 200字推广文案

7. 输出结果格式

**视频创建完成!**

**任务目录**: {路径}
**最终视频**: {文件名}

**播客信息**:
- 标题: {标题}
- 音色: 主播A + 主播B
- 时长: {时长}

---

**内容分析**:

核心金句:...
洞见:...
内容描述(50字):...
朋友圈文案(200字):...

任务目录结构

video-projects/20260322_001/
├── template.mp4          # 复制的模版视频
├── temp/                 # 临时中间文件(可清理)
├── podcast_audio.mp3     # 播客音频
├── {标题}.mp4            # 最终视频
└── 内容分析.txt          # 内容分析

默认配置

配置项 默认值
基础目录 /root/.openclaw/workspace/video-projects
模版视频A v_template.mp4(720×1280,29.6s,上海夜景书房)— 认知/读书/财富类
模版视频B v_template_energy.mov(720×1280,14.7s,能量人)— 能量/潜意识/吸引力类
播客音色1(先说) 主播A(YOUR_SPEAKER_ID_1,在 ListenHub 获取)
播客音色2(后说) 主播B(YOUR_SPEAKER_ID_2,在 ListenHub 获取)
播客模式 quick(3-5分钟)
播客语言 zh(中文)

依赖

  • ffmpeg / ffprobe — 视频处理
  • ListenHub MCP Server — @marswave/listenhub-mcp-server(需要 LISTENHUB_API_KEY)

安装配置

1. 安装 ListenHub MCP

claude mcp add listenhub --env LISTENHUB_API_KEY=\x3C你的API Key> -- npx -y @marswave/listenhub-mcp-server

API Key 在 ListenHub 官网 注册后获取。

2. 配置音色 ID

在使用前,先调用 get_speakers 查询可用音色,将你选择的音色 ID 填入 Skill 配置:

  • YOUR_SPEAKER_ID_1 — 主播A(先说)
  • YOUR_SPEAKER_ID_2 — 主播B(后说)

3. 准备模版视频

将你的背景视频放入 video-projects/ 目录:

  • v_template.mp4 — 默认模版(720×1280 竖屏)
  • v_template_energy.mov — 备选模版(可选)

注意事项

  1. 播客音频生成需要几分钟,耐心等待
  2. 每次任务自动创建日期序号目录(如 20260321_001
  3. 中间文件在 temp/ 子目录,可手动清理
  4. 模板文件复制到任务目录,不影响原始文件
安全使用建议
This skill appears to do what it says (generate podcast-style audio via ListenHub and combine it with templates using ffmpeg), but there are several mismatches you should resolve before installing: 1) The SKILL.md requires LISTENHUB_API_KEY and speaker IDs but the skill metadata does not declare any env vars — expect to provide that API key. 2) The instructions call ffmpeg/ffprobe and execute npx to run @marswave/listenhub-mcp-server; these actions download and run remote code and need your explicit approval. 3) The default BASE_DIR is hardcoded to /root/.openclaw/... which requires write access and may be inappropriate; change it to a safe, user-writable directory. Recommendations: verify the ListenHub package and its npm publisher (@marswave), obtain and store LISTENHUB_API_KEY securely, ensure ffmpeg/ffprobe are installed from trusted sources, and run the skill in an isolated or non-root workspace. If you cannot confirm the npm package or don't want remote installs/executables, decline or request a version that supplies an explicit install spec and properly declares required env vars and binaries.
功能分析
Type: OpenClaw Skill Name: create-podcast-video Version: 1.0.0 The skill automates video creation by executing shell commands (ffmpeg, ffprobe, mkdir, cp) and integrating an external MCP server (@marswave/listenhub-mcp-server) via npx. While the workflow in SKILL.md and references/listenhub-guide.md aligns with the stated purpose of generating podcast videos, the reliance on shell execution and external network-based dependencies constitutes a risky capability profile. Additionally, the ffmpeg command in SKILL.md is potentially vulnerable to shell injection if the AI-generated TITLE variable contains malicious metacharacters, though no evidence of intentional exploitation was found.
能力评估
Purpose & Capability
The skill claims to generate videos from podcast text and to use ListenHub MCP + ffmpeg, which is coherent with its purpose. However the registry metadata declares no required environment variables or binaries, while the SKILL.md explicitly depends on LISTENHUB_API_KEY and ffmpeg/ffprobe and on template files under /root/.openclaw/workspace/video-projects. The undeclared dependencies and hardcoded root workspace path are inconsistent with the stated requirements.
Instruction Scope
Runtime instructions tell the agent to create directories and copy template files under a hardcoded /root path, call ffprobe/ffmpeg, and run an external MCP workflow (create_podcast) that downloads audio. The workflow also instructs running a command that installs an npm package via npx. These steps stay within the skill's stated purpose, but they perform filesystem writes in /root and execute external install commands — operations that exceed a simple, read-only instruction and should have been declared.
Install Mechanism
There is no formal install spec in the registry, yet SKILL.md instructs running: claude mcp add listenhub --env LISTENHUB_API_KEY=... -- npx -y @marswave/listenhub-mcp-server. That invokes npx to fetch and run code from npm at runtime. Using npx to execute a package is effectively downloading/executing remote code and should be declared explicitly; the lack of an install spec is a mismatch and increases risk.
Credentials
The documentation and references require a LISTENHUB_API_KEY and speaker IDs (YOUR_SPEAKER_ID_1/2) but the skill metadata lists no required env vars or primary credential. Requesting an external API key to use ListenHub is proportional to the feature, but failing to declare it is an important omission and a transparency problem. No other unrelated credentials are requested.
Persistence & Privilege
The skill is not marked always:true and does not claim elevated platform privileges. It will create files under /root/.openclaw/workspace/video-projects and instruct installing an MCP server (which may persist components). This is expected for a local video generation workflow but the hardcoded /root path and implicit installation mean it may require elevated write access; consider whether the path should be configurable to a non-root location.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install create-podcast-video
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /create-podcast-video 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
首次发布:双人播客视频生成工具,支持 ListenHub TTS + FFmpeg 合成竖屏短视频
元数据
Slug create-podcast-video
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

CreateVideo - Podcast to Video 是什么?

视频生成工具。当用户说"CreateVideo"、"创建视频"、"生成视频"或提供文案要求制作播客视频时触发。支持双人播客音频生成(通过 ListenHub MCP)、模版视频裁剪合并、内容分析输出。依赖 ffmpeg 和 ListenHub MCP Server。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 86 次。

如何安装 CreateVideo - Podcast to Video?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install create-podcast-video」即可一键安装,无需额外配置。

CreateVideo - Podcast to Video 是免费的吗?

是的,CreateVideo - Podcast to Video 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

CreateVideo - Podcast to Video 支持哪些平台?

CreateVideo - Podcast to Video 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 CreateVideo - Podcast to Video?

由 Conan(@jackyken)开发并维护,当前版本 v1.0.0。

💬 留言讨论