ChatCut says its GPU renderer turns AI code straight into pixels

ChatCut, an AI video editor, says it built a native GPU render engine that turns code written by AI straight into pixels.

The company announced it on X on November 9 and attached a 60-second demo rendered at 4K. The post ends, "Welcome to the real agentic era of motion graphics." That post is still the only public description of the engine. ChatCut's changelog carries no entry about it, and the newest entry there is dated October 30, 2026.

In the post, ChatCut argues that AI motion graphics have mostly been frontend code screen-recorded into video, a method it calls "A clever repurposing, but extremely limiting." It writes that "Quality graphics requires optical effects, particles, 3D lighting and physics simulations, and cinematic post-processing." Its engine, the post says, is one where the AI writes straight to the rendering layer: "So we built a native GPU renderer from scratch, designed for AI to write code directly at the rendering layer." The post sums the approach up as "No frontend frameworks. No Blender. No video generation models. Just code → pixels."

What Is the ChatCut Render Engine?

ChatCut has not named the engine, named the language or graphics API the AI writes, or named the model that writes the code. It has given no release date and no price, and shown no code an outsider can check. The demo is the only evidence so far.

That clip opens on a swirl of bright particles over black, a particle vortex with nothing written on screen. thehype.news's account of the video describes 3D metallic titles reflecting off a glossy floor, a planet glowing at its rim, drifting bubbles and a red and teal nebula, and reports that the clip closes on a card reading "every frame rendered from code." It is a produced demo, the kind of reel that is routinely cherry-picked, and nobody outside ChatCut has run the engine.

The gap the engine is aimed at is real. Most AI motion graphics are made today by an agent writing HTML, CSS or React, a headless browser animating that code, and a tool recording the frames. Remotion, which renders a video from React components, shipped agent skills in January 2026 that let Claude Code, Codex and Cursor write videos, thehype.news reports. HyperFrames, an HTML-to-MP4 renderer that HeyGen built for AI agents and open-sourced in April 2026, works the same way. Both handle titles, charts and kinetic text. Both lean on a browser, which handles particles, volumetric light and physics poorly.

ChatCut already sells motion graphics as editable timeline clips. Its feature page says a described graphic arrives on the timeline as a clip you can trim and restyle, and its documentation says graphics can export as ProRes 4444 files with an alpha channel. A native render engine that draws effects on the GPU is a different layer of the product, and the pages ChatCut sells today do not describe one.

If the engine ships inside ChatCut's editor and its agent plugins, coding agents would gain effects that browser-based tools cannot match yet. That is a conditional doing a lot of work. With no code, no benchmark and no date published, Remotion and HyperFrames stay the tools a creator can use right now.

ChatCut's editor runs in a browser, on macOS and Windows, and inside Claude Code and Codex, with motion graphics on the free plan and paid plans from $21 a month billed yearly; the render engine has no announced date or price.