Notes on using Opus 5.5 for motion graphics
The author says Opus 5.5's generated motion videos are the first that don't feel like pure AI slop, because it skips HTML+CSS and writes canvas drawing code directly, giving control over transitions and shader-based motion blur.
i've been playing with Opus 5.5 for motion graphics (like most people on X) in the past few days, some thoughts: - first time the generated videos don't feel like pure AI slop. the transitions and animations are objectively amazing - Claude used chrome to render the video, but the approach is completely different from Hyperframes and Remotion. it didn't use HTML+CSS to style the videos, but instead wrote code that directly drew on the canvas. that gives the agent full control over the transitions, and allows it to slap some shaders on top for motion blur and other cool effects - what unlocks this approach completely is that Opus 5.5 can handle the layout/elements/etc without any help from a CSS layouting engine. it will literally decide that the letter goes in position X=56, Y=23 and it will look good - another advantage is that none of this requires a browser, in fact you can create a fully GPU accelerated engine to render these videos that can probably run 50x the speed of remotion - i've been experimenting with a lot of approaches to bring motion graphics in @BorumiHQ , and this seems to be the turning point. HTML+CSS was always the wrong abstraction, as HTML+CSS is optimized for web pages, not for complex transitions