Leve Lorenzen, JetBrains and Dave Rogenmoser report different Opus 5.5 video workflows. The model writes animation code; human review and elapsed time vary.
In a September 25 post, Leve Lorenzen shared a promotional video that he says Claude Opus 5.5 helped make. Opus wrote animation, music and sound-effect code, then inspected and fixed frames. This is a way to produce video through code, rather than a dedicated video-generation model.
Lorenzen put his prompt in the first comment. It specifies the product idea, audience, story, a few features, visual style, format and length. The model is asked to write the JavaScript animation, create sound in code, inspect its own frames and improve them. If the brief is incomplete, it must ask questions first. This reads more like a production brief than a one-line request for a polished video.
JetBrains described another route. Its colleague Kristel Cocoli used the HyperFrames toolset, HTML compositions and GSAP animation for a short Kotlin promo. The company says she approved the storyboard, layout sketches and final render in three checkpoints. Its reported 19 seconds refers to rendering the MP4, not the whole production process.
Time also depends heavily on the workflow. Dave Rogenmoser reports using one initial prompt and two small corrections. He estimates about nine minutes of his own time and around three hours of model work. That account cannot be compared directly with JetBrains' 19-second render: the authors measured different stages of different projects.
The product named Echt. in Lorenzen's video is fictional. He mentions TypeSafe AI's Jev as part of the story, but the post does not show a working Jev API integration. For a similar project, his prompt offers concrete requirements for the audience, scenes and duration, plus a check of the rendered frames. JetBrains added three points where a person approved the work.