Six projects across consumer apps, regulated health, and brand film. Every product shot is real app footage or a real store listing — never generated UI. Claim language is reviewed before the script is locked, and every deliverable is reproducible from its source.
Method
Real footage, never generated UI. Image models render on-screen text as gibberish. They are used for people and scenes; the product is always a real screen recording, a real store listing, real screenshots, or a live capture of the web app at a phone viewport. This is the single most common reason a finished video gets thrown away.
Claims reviewed before the script locks. On health, financial or regulated products, restricted language is identified first, and disclaimers are built into editable safe zones so wording can change without a re-render.
Every deliverable is reproducible. Briefs, prompts, source media and the assembly spec are retained together, so any cut can be rebuilt or revised months later without starting over.
Anyone can now generate a video. The difference is being handed the build that regenerates it, being able to prove the words are the words you approved, and being able to say why a pause is 6.20 seconds instead of 0.54.
