How Reelfleur works
Reelfleur is a motion studio for your exact files, not an AI video generator. AI video models synthesize every pixel — which is why logos warp and interface text garbles. Reelfleur inverts the architecture: your files are embedded byte-for-byte, and the AI writes only the code around them — layout, styling, CSS-coded motion.
You upload exact assets
Up to 8 files — app screenshots, logos, icons, product shots (PNG / JPG / SVG, transparency respected). Each asset is tagged by kind: screenshots and photos are never altered; text marks and icons on flat backgrounds may receive light, reversible treatment (like a lightness inversion for contrast) when your brief asks for it.
You pick a style and a size
Six art directions (Editorial, Bold Poster, Neon, Minimal, Playful, Luxe) expressed as background light, shadow character and motion personality — never decorative clutter. Output sizes from 9:16 Story to 16:9 widescreen, chosen before generation because the design is authored for exact pixels.
AI writes code, not pixels
From your one-line brief, the AI writes HTML and CSS that positions, styles and animates your files. Your assets enter the scene as byte-identical images — the model physically cannot repaint them. Built-in skills handle known cases: for example, SVG wordmarks get their real glyph boundaries measured from the vector data so letters can animate individually without ever being sliced.
You refine like a director
“Move the logo left.” “Slower letters.” “Darker background.” Each note edits the current design in place. Coders can open the HTML/CSS blocks directly and hand-tweak for free.
A real render, not a generation
Export runs the scene in a headless browser and captures it frame-by-frame into a crisp H.264 MP4 (30fps, up to 1920px) or PNG still. What renders is exactly what you previewed — the preview and the export are the same code.
The guarantees
- Pixel-exact assets. Screenshots and photos are never filtered, cropped through meaningful content, or redrawn. Ever.
- Verbatim text. On-video copy renders exactly as you typed it — the AI never rewrites your words.
- No hallucinations by construction. Generated scenes run in a sandbox that only permits your uploaded images and inline CSS — there is no mechanism by which invented imagery can enter your video.
- Coded motion, honestly. Animation is CSS — smooth, deterministic, loopable. Reelfleur will not invent camera footage or cinematic worlds. If you need video synthesis, use a video model; if your product must look like your product, use Reelfleur.