MrCarlsama / mrcarlsama-social-transcriber-skill

What is mrcarlsama-social-transcriber-skill?

This is a Python skill for Codex and Claude Code that takes a single Douyin or Xiaohongshu URL and converts it into a complete local content package. You get the original video, extracted audio, SRT subtitles, and a word-level timestamped transcript. It runs locally using faster-whisper for speech recognition, yt-dlp for downloading, and Playwright to handle platform authentication when cookies are needed.

Why is it gaining traction?

The hook is the transcript-first workflow. Researchers and content creators working with Chinese short-video platforms have few tools that produce clean, timestamped text without cloud services. This runs entirely on your machine with a configurable Whisper model size, and it handles the messy reality of platform-specific cookie requirements automatically. The resume capability means interrupted downloads pick up where they left off, which matters when processing longer content.

Who should use this?

Market researchers analyzing Chinese social media, developers building training datasets for speech models, and content teams that need searchable transcripts of short-video content. If you are evaluating Codex or Claude Code as an alternative to GitHub Copilot and want practical skills that solve real workflow gaps, this demonstrates the skill ecosystem in action.

Verdict

At 16 stars with a 0.85% credibility score, this is a niche tool from an individual developer that works but lacks the polish of a mature project. The code is functional and the architecture is sound, but test coverage and documentation are minimal. Worth trying if you actually need Chinese short-video transcription, but do not adopt it as a production dependency without careful validation.