Douyin Copywriter Extractor

Problem Background

Douyin hosts numerous videos with oral content such as tutorials, reviews, or commentaries, which are valuable for content analysis and creative extraction. However, manually transcribing these scripts is time-consuming and inefficient, especially for batch processing. This skill addresses this specific issue by automating the extraction workflow to help users efficiently obtain structured text data from videos.

Core Functionality and Working Principle

The skill is built around four core capabilities, executed in a step-by-step manner:
- Filtering High-Engagement Videos: Automatically filters candidate videos based on configurable thresholds (e.g., min_digg: 20000), prioritizing content with high likes, comments, or shares.
- Downloading Video Files: Uses httpx for asynchronous downloading to local directories, selecting high-definition tracks to ensure quality.
- Script Extraction: Employs a dual-layer strategy—first attempts to fetch subtitle tracks (e.g., VTT/SRT format) from Douyin's API and parses them into plain text; if subtitles are missing or fail, it falls back to local ASR with faster-whisper, requiring FFmpeg for audio extraction.
- Information Retrieval: Supports enhanced retrieval of video interaction data and subtitle information via Chrome MCP, providing a fallback when script-based methods are restricted.

The key workflow involves: users submit video IDs or links; the skill parses and validates the data; filters videos based on configuration; downloads and extracts scripts, outputting to a specified directory (e.g., ~/抖音下载/) with files named by search keywords (format yyyy-MM-dd keyword.txt), where the script content includes timestamps and text.

Applicable Scenarios and Considerations

This skill is suited for scenarios requiring batch analysis of Douyin video scripts, such as market research, competitive content monitoring, or creative inspiration collection. However, note the following boundaries:
- Videos must contain oral content (e.g., commentary or interviews); silent or music-only videos cannot yield scripts.
- API subtitle extraction only works for videos with embedded subtitles; some videos may lack subtitles, in which case the skill automatically falls back to the ASR approach.
- ASR functionality requires installing FFmpeg and faster-whisper, with an initial model download (around 75MB); ensure stable network connectivity.
- High-volume downloads may trigger Douyin's anti-crawling mechanisms; it is advisable to control processing frequency (e.g., process_count: 1) and set appropriate timeouts.
- Network instability or API changes might cause extraction failures; retries or switching to the Chrome MCP method for data retrieval are recommended as alternatives.