
HyperFrames
✓ Editorially verifiedOpen-source video composition framework that lets AI agents build videos by writing HTML, CSS, and JS.
In short
HyperFrames is a free, open-source framework letting AI agents build videos using HTML, CSS, and JS. It offers deterministic, code-based video composition for developers.
Pick HyperFrames if you want an AI coding agent to compose and iterate on videos as code rather than driving a timeline UI.
Skip it if you want a click-and-drag video editor or one-shot text-to-video generation without touching markup.
HyperFrames is an open-source framework from HeyGen that turns video editing into a code task AI agents can actually perform. Instead of wrapping a timeline UI around a model, it asks agents to compose scenes as HTML, CSS, and JavaScript, with a catalog of pre-built blocks (transitions, VFX, device mockups, lower-thirds) you can import and remix. It ships as an NPM package and as a Claude Code skill installable via `npx skills add heygen-com/hyperframes`.
The target user is a developer pairing with an AI agent who wants programmatic video generation without leaving the terminal. Because everything renders from web standards, output is deterministic, diffable, and easy for an LLM to iterate on, three things traditional NLE-driven AI tools struggle with. The framework itself is Apache 2.0 and free; HeyGen is presumably betting it funnels users into its hosted avatar/video products downstream.
There's a public playground at hyperframes.dev, a quickstart, and a browsable block catalog. It's early — the marketing-promo demo and a handful of VFX blocks are what's on display — so expect to write your own components for anything beyond the included primitives.
The smartest take on agentic video we've seen this year: treat the composition as a web document, hand it to Claude Code, and iterate. It's early and the block library is thin, but the architecture is right and HeyGen has the distribution to make it stick.
— The AI Tool Bible editorial team
Pros
- ✅ Open-source under Apache 2.0 with a public GitHub repo
- ✅ Ships as a first-class Claude Code skill, not a vague API
- ✅ Web-standard output (HTML/CSS/JS) is diffable and agent-friendly
- ✅ Block catalog accelerates common compositions like device mockups and VFX
- ✅ Backed by HeyGen, which actually ships video products
Cons
- ⚠️ Requires writing or generating code; not a no-code editor
- ⚠️ Block catalog is still small and skewed to marketing promos
- ⚠️ Newer project with limited real-world case studies
- ⚠️ Render pipeline and hosting story aren't spelled out on the landing page
Use cases
Frequently asked
- How much does HyperFrames cost?
- HyperFrames is completely free and open-source under the Apache 2.0 license. There are no subscription fees mentioned, though HeyGen may funnel users toward its other hosted products.
- What AI models does HyperFrames support?
- It uses a multi-model approach, specifically designed as a Claude Code skill. It supports Bring Your Own Agent (BYO agent) workflows, allowing you to use your preferred AI coding agent.
- Is HyperFrames suitable for non-developers?
- No. It is designed for developers pairing with AI agents who want programmatic video generation. If you prefer a click-and-drag timeline editor or one-shot text-to-video without touching markup, you should skip it.
- How do I install HyperFrames?
- It ships as an NPM package. You can also install it as a Claude Code skill using the command `npx skills add heygen-com/hyperframes` to integrate it into your terminal workflow.
- What features are currently available?
- It includes a catalog of pre-built blocks like transitions, VFX, device mockups, and lower-thirds. However, it is early-stage, so you may need to write custom components for anything beyond these included primitives.
Explore related
Compare with similar tools
All in Video →Runway
FeaturedPro-grade AI video editor and Gen-4 generation.
Sora
FeaturedOpenAI's flagship text-to-video model.
Luma Dream Machine
Fast, accessible text-to-video with strong camera control.
HeyGen
Avatar video + lip-sync translation at scale.
Google Veo
Google DeepMind's flagship text-to-video model with native audio generation and cinematic camera control.
Higgsfield
AI video and image generation suite that aggregates 30+ frontier models under one workflow.