Omma's AI canvas now accepts video and 3D files as inputs
By Steven Van ·
Creators can feed real GLB, GLTF or MP4 assets into Omma to guide generations or drop them straight into a deployable page.
Omma, the AI canvas from Spline, just closed one of the bigger gaps between generative building and real-world design assets: on June 30, 2026 the team announced that the Canvas now accepts video and 3D files as inputs for in-canvas generations. Instead of describing a 3D model or a background clip and hoping the AI approximates it, you can now drop the real asset in and build around it.
Real 3D models and background video, live on the page
The headline use case is deployment. You can now build and ship sites and portfolios that feature real-time 3D models and background videos directly in the layout — not static renders or placeholder loops. Because Omma outputs production-ready, deployable web experiences rather than mockups, a GLB or GLTF model you feed in can end up as an interactive element on a live page, and an MP4 can sit behind your hero section as motion. For designers who have wrestled with hand-wiring WebGL scenes or fussing with video backgrounds in code, that is a meaningful shortcut.
Two ways to use an asset: guidance or content
The update splits input usage into two modes. First, a 3D file or video can act as guidance — steering a generation the way a reference image does, so the AI matches the shape, motion, or style of what you provided. Second, the same file can be dropped in as a finished asset that lives inside your generation as-is. That distinction matters: you can use a rough model to direct a scene, or place a polished, client-approved model straight onto the canvas without regenerating it.
Fitting into Omma's parallel-agent pipeline
Omma already runs multiple AI agents at once — one handling code generation, another building 3D meshes, another producing images — all working from a single prompt. Adding video and 3D as accepted inputs extends that ingestion layer, which reportedly handles formats from CSV, JSON, and DOC to GLB, OBJ, GLTF, PNG, SVG, and MP4. Launched by Spline in March 2026 and already used across a design community of millions, Omma is positioning itself between heavy 3D authoring tools and code-first web generators. Letting creators bring their own motion and geometry into that flow reinforces the pitch: describe it, or supply it, and build the whole thing in one canvas.
Sources: Abduzeedo, Spline changelog, Omma.