@yaohui12138: Holy crap, isn't this basically the legal money printer of the AI era??? Replicate any!! Any!! Viral ad, video, street …

Holy crap, isn't this basically the legal money printer of the AI era??? Replicate any!! Any!! Viral ad, video, street interview... etc., in just a few minutes Swap out a face, swap the topic, and batch-produce 100 videos of the same viral style?? You're killing off all the directors slaving away over topic research and scripts—what are they supposed to do, just die? And the project is open-source too—what kind of cyber bodhisattva is this? 😭 Let me first explain what this thing actually does— It teaches your Coding Agent (Claude Code, Codex...) how to make videos in the truest sense Not that "input a sentence and generate a blurry AI video" nonsense—it's the real deal: complete, finished videos with scripts, voiceovers, subtitles, B-roll, special effects And the core capability is: Clone! (Cloning) Toss in a viral video, run one command, and it breaks down the entire structure of that video —How the shots cut, how the lines are written, what subtitle style, where to insert B-roll, how to trigger effects, how to arrange the pacing—all turned into editable components Then you can swap 'em out like skins: Swap the host from a real person to a banana cat, the topic from soccer to tech scene, the language from English to Chinese, the product from creatine to skincare Same viral structure, endless reskins, batch production!! Let me first say how painful the traditional way is To replicate a 20-second viral short video, the normal process is like this: Watch the original video, manually break down the structure → Half a day writing scripts, tweaking lines → Half a day sourcing materials, generating AI visuals → 1-2 days voiceover, adding subtitles, doing effects → 1 day editing, adjusting pacing, rendering → Half a day Back-and-forth revisions → Add another two days One person doing it all, fastest is 3-5 days per video; if you're an MCN batch-producing, one editor can only crank out 3-5 a day And the killer part: Every one starts from scratch, the viral structure's in your head, but you can't reuse it, can't automate it I gave it a try I grabbed a random product promo ad, tossed it into Claude Code, input one command: /hypit clone this video Ten-odd minutes later, the finished piece was out. Script, voiceover, subtitles, B-roll, effects—all auto-generated, ready to use right away Then I did something even more absurd—I said: Swap the character It only regenerated the character-related materials, reused everything else, and spat out another one in minutes Same ad, different character, boom—a new ad's born Two videos, same viral structure, different skins, cost is just the gen for images and video!! What does this mean? As someone who's built products before, what I see isn't just "can replicate videos"—it's an entire content production line getting rebuilt: 1. Making AI short videos: Spot a viral one, clone it, swap the persona and topic—now it's yours, crank out dozens a day 2. MCNs and content teams: Same workflow spits out versions in different languages, formats, CTAs—one person does a whole team's work 3. Knowledge paywall stuff: Course videos, lead-gen videos—define a template, batch-generate variants 4. Taking ad gigs: Client hands you a reference video, clone it and tweak—deliver. What used to quote for thousands now costs pennies Short video production used to be craftsmanship, relying on editors' experience and time Now it's engineered—viral structures can be coded, modularized, version-controlled, batch-copied You might ask: How's this different from other AI video tools? A few key points: It's not a video generation model Seedance, Sora—these generate materials, but "generating materials" ≠ "completing a video" A full video still needs script, voiceover, subtitles, editing, B-roll, effects, arrangement, rendering This project handles that whole backend chain It's not a traditional editing tool—no dragging timelines, code controls the video. All elements aren't nailed to a specific second on the timeline—they're tied to the corresponding script line—so tweak one line, the whole segment updates It's not one-click template generation. You can tweak any component, even build your own workflow from scratch Plus, it's open-source, free, full code on GitHub The project's called Hypit TL;DR: An open-source language and compiler system for AI Agents to make videos Clone any viral video with AI agents Don't forget to head to GitHub and drop it a Star—after all, this is straight-up cyber bodhisattva energy, we reap the rewards, least we can do is leave a star! 👇