douyin-video-to-doc

name	douyin-video-to-doc
description	Use when the user provides a Douyin link or share text and wants one-document-per-video output with transcript, source metadata, and a detailed content writeup rather than a short summary

Douyin Video To Doc

Turn a Douyin share link into a reusable document bundle:

• source video metadata
• local video/audio files
• local transcript files
• one Markdown document per video

Use this skill when the user wants to archive, study, summarize, or structurally organize the full content of a Douyin video.

Output Rule

Create one Markdown file per video under the configured output directory.

Default output directory:

%USERPROFILE%\douyin-video-docs

Override with:

DOUYIN_VIDEO_TO_DOC_OUTPUT_DIR

Keep the global skill itself under:

C:\Users\dolan\.codex\skills\douyin-video-to-doc

On this machine, if you want outputs under your project workspace root, set:

$env:DOUYIN_VIDEO_TO_DOC_OUTPUT_DIR = "D:\dolan_env\temp\project\personal\douyin-video-docs"

Workflow

1. Run the helper script to fetch metadata, download media, extract audio, and generate transcript files.
2. Inspect the transcript and correct obvious ASR mistakes through context.
3. Write a final Markdown document with:
   • basic video info
   • processing notes
   • detailed content writeup
   • transcript raw text
4. Be explicit when transcript text came from local ASR rather than official captions.

Script

Use:

python C:\Users\dolan\.codex\skills\douyin-video-to-doc\scripts\douyin_video_to_doc.py --input "抖音分享文本或链接"

Optional flags:

python C:\Users\dolan\.codex\skills\douyin-video-to-doc\scripts\douyin_video_to_doc.py --input "..." --model small --output-dir "D:\dolan_env\temp\project\personal\douyin-video-docs"

The script produces:

• VIDEO_ID.mp4
• VIDEO_ID.wav
• VIDEO_ID.<model>.srt
• VIDEO_ID.<model>.txt
• VIDEO_ID.metadata.json
• YYYY-MM-DD-douyin-<slug>-<video_id>.draft.md

Example:

python C:\Users\dolan\.codex\skills\douyin-video-to-doc\scripts\douyin_video_to_doc.py --input "7.99 复制打开抖音，看看【每日AI评论的作品】简单讲讲最近最火的Hermes Agent https://v.douyin.com/j6ySf9z4GKg/"

If you want to pin output to a custom directory for the current shell:

$env:DOUYIN_VIDEO_TO_DOC_OUTPUT_DIR = "D:\dolan_env\temp\project\personal\douyin-video-docs"
python C:\Users\dolan\.codex\skills\douyin-video-to-doc\scripts\douyin_video_to_doc.py --input "https://v.douyin.com/xxxx/"

Writing Standard

The final document should not stop at a framework-level summary.

Expand the video into:

• the speaker's main claim
• argument structure
• implementation details
• examples and mechanisms
• assumptions and caveats
• inferred corrections for obvious ASR errors

If a key proper noun is uncertain, say it is inferred from context.

Verification

Before claiming completion:

• confirm the draft and transcript files exist
• read the transcript text
• make sure the final Markdown includes both transcript and detailed writeup

Notes

• Douyin pages often do not expose reusable official subtitles.
• In that case the transcript is local ASR output, not platform-native captions.
• Prefer the small model when quality matters more than speed.