Hey Product Hunt 👋 I make demo videos for my own product, and the part I hated wasn't recording — it was the hour afterwards, in a timeline, placing zoom keyframes on top of my own clicks. The information was already there. I'd clicked the button. Why was I telling an editor where I'd clicked? So Phtonix Studio does the editing pass at record time. A tiny in-page tracker reports the *coordinates* of each click and keystroke — never the characters — and the renderer turns those into camera moves that are baked into the file. Typing was the hard part: chasing the caret letter by letter looks awful, so it opens tight on the first few characters, holds still, and reframes only when the caret is about to leave the shot. Two things I decided not to compromise on: 1. Constant frame rate. It seeks each output frame and encodes it through WebCodecs, so 60fps means exactly round(duration × 60) frames — not "60fps-ish with drift." 2. Nothing leaves your machine. Not a policy promise — the manifest sets connect-src to 'none', so the extension is physically incapable of making a network request. There's no server, no account, no analytics. Screenshot mode came from the same itch: every framing tool hands you eleven presets. Here you pick both gradient colours yourself, crop at full resolution, and the 3D tilt is a real projection — the far edge foreshortens and the shadow follows the tilted silhouette. Known gaps, since you'll find them anyway: export is WebM (MP4 is next), there's no annotation/mark-up yet, and it can only zoom inside web pages — desktop apps record fine but stay at full frame. Free, no card. Would love to know where the zoom guesses wrong on your product.