# Real2Simulation

- **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:** [Koushik Sadhak Jeet](https://cerebralvalley.ai/u/I-am-Mr-Rookie)
- **GitHub:** https://github.com/I-am-Mr-Rookie/real2simulation
- **Demo video:** https://www.youtube.com/watch?v=9E52-Bpz3gg
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3346

Real2Simulation turns a phone video into real physics you can measure, replay in 3D and change.

The problem: In a school lab, students film an experiment, but getting real numbers from the video means clicking the object frame by frame. A simulation is a separate tool that never matches what really happened. Most AI tools also just give a number with no proof and no uncertainty.

What it does:
1. Film an experiment. You upload a phone clip and tell the agent what to measure, for example "Measure the falling book, it is 20.3 cm wide". It tracks the object, fits the motion and gives a result with uncertainty. Our phone book drop measured 9.07 +/- 0.62 m/s^2. The demo video shows another drop of the same book: 9.1 +/- 1.1 m/s^2. Over 12 drops of the same book the median was 9.03: repeatable, but still a little low, and we say that openly.
2. Redo it with changes. The measured points stay fixed in a 3D replay while you change things like gravity (Earth to Moon) and watch the simulation move away from reality.
3. Film a room. Your room video becomes an editable 3D physics lab. You can add a ball, a ramp or a pendulum, press Play and run experiments inside your own room.
4. Use it in a class, with no account needed. A ready physics lab template with guided lessons ("Does a heavier ball fall faster?"), Fit trace with uncertainty, and a Guess the Gravity game.

Honest by design: Every number comes from tracked data, the fitter or the simulator, never written by the AI. Results are always labelled MEASURED, SIMULATED or MODEL-ESTIMATED, with "scale assumed" or "scale unverified" where needed. If the agent is missing a size or a frame rate, it asks you instead of guessing.

Live: https://real2simulation.rookiecoders.tech

## More from Meta Global AI Developer Hackathon

- [Yarrow](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3343)
- [Sprout](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3344)
- [Timo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3345)
- [Hypa Intelligence](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3347)
- [almona.eg](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3348)
- [Derm Forge](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3349)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3346. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt