douyindownloadwhisper

Windows环境下抖音无水印视频下载和本地Whisper语音转文字及自动语义分段文案提取工具。

Security Scan

The documentation and registry claim a local-only Whisper + built-in-LLM tool with no required credentials, but the included code reaches out to third‑party APIs and references API keys that are not declared—this mismatch is a red flag.

Purpose & Capability

The skill's stated purpose (download Douyin videos, local Whisper transcription, and OpenClaw-built-in LLM segmentation) is plausible, but the code includes calls to external services (api.siliconflow.cn and api.minimaxi.com) and mentions environment variables (SILI_FLOW_API_KEY, MINIMAX_API_KEY) that are not declared anywhere in the SKILL.md or registry metadata. The presence of external-API integration is not explained by the description and is disproportionate to the 'local-only' claim.

Instruction Scope

SKILL.md instructs local execution (Whisper and ffmpeg) and says segmentation uses the OpenClaw built-in LLM, but the code actually invokes external endpoints (MiniMax via curl, likely SiliconFlow) and will upload/POST data (audio/text) to third parties if API keys are present. The runtime docs omit these network behaviors and the environment variables the code expects. There are also odd/malformed example paths in SKILL.md (missing path separators) that may cause unexpected behavior.

Install Mechanism

There is no install spec (instruction-only), which is lower risk; however, the code itself spawns external programs (ffmpeg, whisper, curl) and will automatically trigger network activity (Whisper may auto-download models, the script calls external APIs). Lack of an install step reduces installer risk but does not prevent the script from performing network I/O at runtime.

Credentials

Registry and SKILL.md state no required environment variables, but the code uses and documents SILI_FLOW_API_KEY and MINIMAX_API_KEY (plus behavior that falls back to process.env). Requiring API keys for third‑party transcription/segmentation is not justified in the public docs and increases the risk of sensitive data (audio/transcripts) being sent to external services.

Persistence & Privilege

The skill is not marked always:true and does not request permanent platform privileges. It appears to be a normal, user-invoked skill; no evidence it modifies other skills or system-wide configs was found in the provided portion of code.

What to consider before installing

Do not install or run this skill without understanding and consenting to its network behavior. Specific points to consider before using: (1) the documentation claims local-only processing, but the code will contact api.siliconflow.cn and api.minimaxi.com and can upload audio/text—this is not documented and requires API keys. (2) If you must use it, inspect the entire douyin.js file end-to-end (the snippet is truncated) to find all endpoints and data flows, and confirm what is sent. (3) If you want purely local processing, modify the script to remove external API calls (delete or disable SiliconFlow/MiniMax calls and the related env-var checks) and ensure Whisper runs fully locally. (4) Run the script in an isolated environment (VM or sandbox) and monitor outbound network connections before providing any API keys or running on sensitive data. (5) The SKILL.md also contains malformed example paths—verify and correct them to avoid accidental behavior. If the author can confirm the code will not send audio/text to third parties and update SKILL.md to declare required env vars and endpoints, the assessment could change to benign.

一、技能基础信息

name: douyin-download-whisper description: 抖音无水印视频下载和文案提取工具，适配Windows系统，支持无水印视频解析下载、语音文案提取（锁定Whisper base模型）及文案语义分段。 metadata: openclaw: emoji: 🎵 requires: bins: [ffmpeg, whisper] config: whisper_model: base # 锁定Whisper base模型，确保转写效率与兼容性，适配Windows本地运行

二、技能介绍

douyin-download 是一款适配 Windows 系统的抖音辅助工具，核心功能为无水印视频下载、视频文案提取（含语音转文字）及文案语义分段，依托本地 Whisper 实现语音转写，调用 OpenClaw 内置 LLM 完成语义分段，操作简单、可直接通过命令行调用。

三、核心功能

• 🎬 获取无水印视频下载链接：解析抖音分享链接，提取无水印视频源地址

• 📥 下载抖音视频：将解析后的无水印视频直接下载至指定目录

• 🎙️ 语音文案提取：通过本地 Whisper 工具，将视频中的语音内容转写为文字文案

• ✂️ 文案语义分段：自动调用 OpenClaw 内置 LLM，对提取的文案进行自然语义分段，提升可读性

四、环境依赖（Windows 系统必配）

• Windows 安装（推荐使用阿里云镜像）

  pip install -i https://mirrors.aliyun.com/pypi/simple/ --upgrade openai-whisper
  

该技能依赖以下工具，需提前安装并配置，否则无法正常运行：

1. ffmpeg：用于视频解析与处理，需安装后添加至 Windows 系统环境变量 PATH（安装后可在命令行输入 ffmpeg -version 验证是否配置成功）。

2. whisper：用于语音转文字，通过 Python 命令安装，安装命令：pip install -U openai-whisper，安装后可在命令行输入 whisper --version 验证是否可正常调用。

3. Node.js：需提前安装 Node.js 环境，确保 node 命令可在命令行正常使用（用于执行 JS 脚本）。

五、环境变量

（此技能无需配置额外环境变量，仅需确保上述依赖工具正常安装并配置完成即可）

六、使用方法（Windows 命令行专用）

所有命令均需在 Windows 命令提示符（CMD）或 PowerShell 中执行，复制命令后替换「抖音分享链接」为实际链接即可使用，路径可根据需求自行修改。

6.1 获取视频信息

功能：解析抖音链接，获取无水印视频地址、视频标题等基础信息。

node "%USERPROFILE%.openclawworkspaceskillsdouyin-downloaddouyin.js" info "抖音分享链接"

6.2 下载视频

功能：将无水印视频下载至指定目录，默认下载至 C:Tempdouyin-download，可通过 -o 参数修改下载路径。

node "%USERPROFILE%.openclawworkspaceskillsdouyin-downloaddouyin.js" download "抖音链接" -o C:Tempdouyin-download

说明：若指定的下载目录不存在，工具会自动创建该目录。

6.3 提取文案（自动语义分段）

功能：下载视频并提取语音文案，自动进行语义分段，生成可读性强的文字内容。

# 如需使用本地 Whisper，请确保已安装 whisper 可执行文件（安装命令：pip install -U openai-whisper）
node "%USERPROFILE%.openclawworkspaceskillsdouyin-downloaddouyin.js" extract "抖音链接"

• 核心逻辑：先下载视频 → 提取视频中的语音 → 通过本地 Whisper（已锁定base模型）转写为文字 → 调用 OpenClaw 内置 LLM 进行自然语义分段，base模型体积小、运行高效，适配Windows本地环境。

• 注意：首次使用 Whisper 时，会自动下载锁定的base模型，体积较小，无需等待过长时间，需保持网络通畅。

6.4 跳过语义分段

功能：提取文案但不进行语义分段，直接输出完整的转写文字。

node "%USERPROFILE%.openclawworkspaceskillsdouyin-downloaddouyin.js" extract "抖音链接" --no-segment

七、常见问题

• 问题1：执行命令时提示「node 不是内部或外部命令」？ 解决：未安装 Node.js 或未将 Node.js 添加至系统环境变量，重新安装 Node.js 并重启命令行。

• 问题2：提取文案时提示「whisper 不是内部或外部命令」？ 解决：Whisper 安装未成功或未添加至环境变量，重新执行 pip install -U openai-whisper，并检查 Python 的 Scripts 目录是否在系统 PATH 中。

• 问题3：下载视频失败？ 解决：检查抖音链接是否有效（需为完整的抖音分享链接），网络是否通畅，目标下载目录是否有写入权限。

八、注意事项

1. 本工具仅用于个人学习、研究使用，请勿用于商业用途或下载侵权视频，遵守抖音平台规则。

2. 若抖音平台接口更新，可能导致工具解析失败，需等待技能更新适配。

3. Whisper 转写精度受视频语音清晰度影响，模糊语音可能出现转写误差。