louisedesadeleer/winning-structures — explained in plain English
Analysis updated 2026-05-18
Paste a viral short video link and see its script broken into structural beats.
Rewrite a viral script line by line in your own words while keeping the original pacing.
Record a new video using the built-in teleprompter timed to match the original creator.
Read the structure-anatomy notes to study why certain videos retain viewer attention.
| louisedesadeleer/winning-structures | aredotna/api-examples | ceelog/openweread | |
|---|---|---|---|
| Stars | 15 | 15 | 15 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Browsing the library needs only Python, but analyzing new videos requires installing yt-dlp, ffmpeg, Whisper, and Claude access.
Winning Structures is a tool for content creators who want to reuse the pacing and structure of a viral short video, like a TikTok, Reels, or YouTube Shorts clip, while writing their own words. You paste in a link to a video, and the app downloads it, transcribes every word along with its exact timing, and then asks Claude to break the script into structural beats, such as the hook, the main points, the key line, and the call to action. The idea behind the project is that videos do not go viral just because of editing, but because every sentence in the script is doing a specific job, and a document included in the repository explains that theory in more depth. Once a video is analyzed, you can open it and rewrite each line with your own words, and the app estimates how long your new line will take to say based on the original creator's speaking pace. You can then play the whole rewritten script back as a teleprompter while recording yourself, with a mirror mode, adjustable speed, and a countdown for each beat. Running the app just to browse an existing library and use the teleprompter needs nothing beyond Python, since the frontend is already built and included. Analyzing brand new videos requires a few additional command line tools to be installed separately: yt-dlp to download the video, ffmpeg to handle audio, Whisper to transcribe speech, and either Claude Code or an Anthropic API key to do the beat mapping. Analyzing one video takes roughly one to two minutes, and the downloaded video file itself is deleted afterward, keeping only the thumbnail and the extracted structure. The project is released under the MIT license.
Winning Structures lets you copy the beat-by-beat pacing of a viral short video and rewrite it with your own words using a teleprompter.
Mainly TypeScript. The stack also includes TypeScript, Python, React.
MIT license, use freely for any purpose including commercial use as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.