----------------
🎥 Video
===================
HyperFrames: writing video as code, rendered through Claude
A new tutorial from the Israeli AI blog LetsAI walks through a workflow where video production starts as a conversation with Claude instead of a timeline editor. The clip covered there was built entirely with Claude: the model computes each frame as code, and a render engine captures the frames in sequence to produce a finished MP4. The tutorial's framing borrows from Eadweard Muybridge's 1878 horse motion studies: motion is a sequence of still frames, and whoever controls each frame controls the motion.
What HyperFrames is
HyperFrames is a free, open-source project by HeyGen. It represents videos as HTML compositions and renders them to video files. It ships in two parts:
• A command-line tool, invoked as npx hyperframes, which downloads automatically on first run.
• A set of Skills for Claude Code, installed with npx skills add heygen-com/hyperframes. Skills are instruction files Claude reads on demand; the HyperFrames ones teach it how to structure a valid composition, how to schedule animations on the timeline, and how to test and render the result.
Two content types, two workflows
The tutorial separates two use cases:
1. Animation from scratch: ads, teasers, and explainer videos, built entirely as code.
2. Editing footage filmed by the user: trimming silences, Hebrew subtitles, zoom effects, and converting landscape footage to vertical format for social.
In both cases the output is an MP4, and the work happens in natural language with approvals at each step.
Two ways to run it
• Connector mode: HyperFrames plugs into Claude as a connector, with nothing to install. Lower ceiling on control, faster entry point.
• Claude Code mode: runs locally on your machine with the Skills installed. The tutorial claims benefits here in precise Hebrew text rendering, working with your own footage files, and local rendering rather than anything cloud-side.
Why it matters
For teams producing short social video, the interesting part is not the gimmick but reproducibility: a video defined in HTML is versionable and diffable, and revisions become prompt-driven edits to specific scenes, in a way a traditional editor project is not. Localization also runs through the same pipeline, since subtitle generation and text rendering are part of the composition. The tutorial specifically highlights accurate Hebrew rendering as a differentiator in the local setup, which is relevant wherever RTL text is a recurring pain point in automated video tools.
Caveats
This is a write-up of someone else's workflow, not a tested one. The source is an AI training and consulting blog, so the tone is partly promotional; rendering quality, Hebrew font handling, and render times deserve personal verification before committing a real project to this. HeyGen's own prompting documentation is available for review.
References
• Documentation: hyperframes.heygen.com
• Repository: github.com/heygen-com/hyperframes
• Official prompting guide: hyperframes.heygen.com/prompting/overview
🔹 HyperFrames #ClaudeCode #AI #VideoGeneration #tool
🔗 Source: https://letsai.co.il/hyperframes/?ref=8
