# teardown

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Walid Elonok](https://cerebralvalley.ai/u/itswal)
- **GitHub:** https://github.com/walsoup/teardown
- **Demo video:** https://drive.google.com/file/d/1g9hrh9aCRXq6NVVrkdRiyCaDvcxZ6YXu/view?usp=drivesdk&authuser=walidelonk1234%40gmail.com
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4120

Repair videos contain useful knowledge, but finding the exact moment, part, or action often means scrubbing through a long recording. teardown turns narrated repair footage into a step-by-step visual guide: a keyframe per step, the relevant part highlighted when segmentation finds it, zoom insets, tools and parts, and a searchable step index. The web app accepts an uploaded video or a YouTube link and lets you jump between steps and the source footage.

The pipeline combines transcription in 30-second windows, a model-generated plan, a mechanical check that each step's anchor quote occurs in the transcript, and part segmentation on one keyframe per step. That keyframe-first approach keeps segmentation costs down. The recorded demo generated 21 steps from a TheCod3r Xbox Series X repair video used with permission, at about $0.11 in API usage for that run.

The guide is explicit about its limits: timestamps are approximate (+/-30 seconds), missed part matches are labeled "not highlighted", and output is a viewing aid rather than validated repair instructions. Browser-scoped guides auto-delete after three hours. The public demo has a metered budget and rate limits. YouTube ingest currently depends on a volunteer proxy and is not guaranteed.

Live app: https://teardown-0k03.onrender.com/

## More from Meta Global AI Developer Hackathon

- [policy-evidence-engine](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4117)
- [Kaung Zin Hein](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4118)
- [Moltbot Den](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4119)
- [HireSense by LV8N Tech](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4121)
- [Team Crew Daniel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4122)
- [PMO (Permanent Memory Optimizer)](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4123)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4120. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt