# Codekrafters - **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:** [Ishaan Gupta](https://cerebralvalley.ai/u/Ishaan-GPT), [Vignesh B S](https://cerebralvalley.ai/u/vigneshbs) - **GitHub:** https://github.com/Ishaan-Gpt/metahack - **Demo video:** https://youtu.be/AtYfpaLwLOc - **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery - **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4860 The next era of video will be written in code, not painted by a model. Distribution now decides which products, creators and ideas win, and distribution means short-form video, every single day. The bottleneck was never the filming. It is the editing, the craft that turns a flat take into something people stop scrolling for, and that craft is slow, expensive and locked inside the heads of professional editors. Generative video tried to skip the craft and hands back a synthetic face saying words you never said, so the people who need video most trust it least. Edit Easy is a code-native editing engine. Drop in the raw take you already filmed and get back the cut a top short-form editor would make: reframed for 9:16 around your face, dead air gone, kinetic captions with one lit hero word per line, photo splits, list builders, cards, words that sit behind your head, motivated transitions, sound design on every change and a graded brand frame. Every frame is drawn on purpose, and your real face and voice are in every one of them. The hard part was turning taste into something an engine can execute. We reverse-engineered professional before-and-after edits frame by frame, measuring type, colour, timing and the rhythm of change, and rebuilt them 1:1 in code beside the originals. From that we distilled an edit grammar of beat roles, layouts, subject treatments, text systems, transitions and sound events, guarded by hard rules no edit may break: text never covers a face, never a dead frame, never the same transition twice in a row. Meta’s Muse models sit exactly where judgment lives. Muse Voice Transcribe hears the take, Muse Spark reads all of it once and directs every beat under a strict schema, and a deterministic engine checks and renders each decision. The model decides and the code draws, so every edit is consistent, repeatable, cheap, and can never invent a face. Then the edit stays alive. It plays instantly as a live preview in the browser, every caption, photo and effect is clickable, and you change any of them by asking in plain words, matched against an edit dictionary so refinements are instant and free. Render the final MP4 with a before/after split that sells the difference in one second. Two engines ship today, Card Studio and Creator Mix, with a library of effects chosen per video by relevance, and the same director can drive any new engine built from another editor’s style. Where this goes is a library of editing dialects anyone can wear, turning the most expensive skill in content into something every creator owns ## More from Meta Global AI Developer Hackathon - [LifeLens](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4857) - [CleanQuest](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4858) - [LENHROS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4859) - [Cortivra](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4861) - [Solo - Sohail Mughal](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4862) - [CairoFlow](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4863) --- Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4860. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt