Short Video Copy Extract Pro

Problem

Short-video copy teardowns often stall in three places: platform description and transcript may come from inconsistent sources, headless browsers can be blocked by anti-scraping, and extracted copy still lacks structure judgment for opening hooks, information density, emotional tension, and CTA effectiveness. This skill chains extraction, analysis, and rewriting into one Python pipeline, reducing manual rework and repeated cleanup.

How It Works

The script first uses DouSnap.com to fetch title, description, transcript, and media links; on failure, it falls back through HelloTik.app, yt-dlp, and Whisper. The text then enters the analysis engine, parses a five-part structure, identifies hook, setup, core, turn, and CTA, and calculates a 0-100 weighted score. It outputs the original, readable, and condensed versions, preserves core information, only fixes punctuation, fluency, filler removal, and compression, and supports MD, JSON, CSV, and plain-text export.

Limits

Transcripts may be audio ASR content, so videos without narration may contain BGM lyrics; compare them with description before using them. dousnap depends on browser DOM and works best with CDP connected to real Chrome; without CDP it may return only media links. If description, transcript, and local ASR are all missing, it returns empty results rather than inventing copy. Long-video and batch runs also need attention to timeout, platform support, privacy risk, and reproducible teardown fields.