Marketing skills for Claude Code and AI agents. CRO, copywriting, SEO, analytics, and growth engineering.
Hyperframes
Open-source programmatic video framework from HeyGen. Create videos from HTML/CSS/JS — no React, no proprietary DSL. Designed for AI agent workflows.
Capabilities
| Integration | Available | Notes |
|---|---|---|
| API | - | Library, not a hosted service |
| MCP | - | - |
| CLI | Yes | npx hyperframes render |
| SDK | Yes | @hyperframes/producer for rendering from code; the hyperframes package itself is CLI-only |
Why Hyperframes
- LLM-native: AI models generate better HTML than React components — plain web standards, no framework DSL
- Deterministic: Same input always produces identical output (ideal for automation)
- Open source: Apache 2.0 license, zero per-render fees
- Agent-friendly: Any coding agent that can write HTML can create videos
Install
npx hyperframes <command> # or: npm install -g hyperframes
Requires Node.js 22+ and FFmpeg. The hyperframes package is a CLI only; it has no importable API. For rendering from your own code, use @hyperframes/producer (below). Checked against v0.8.114, Oct 2026.
Quick Start (CLI)
npx hyperframes init my-video # scaffold a project from a template
cd my-video
npx hyperframes preview # live preview in the Studio (localhost:3002)
npx hyperframes render -o output.mp4 # render the project's index.html
npx hyperframes render -c ./promo.html -o promo.mp4 # render a specific composition
npx hyperframes lint . # catch composition errors before rendering
How a Composition Works
A composition is an HTML file. The root element declares the canvas and total length, each visible element is a clip with its own start, duration, and track, and animation runs on a paused GSAP timeline that Hyperframes drives frame by frame. This is the shape of the package's own blank template:
<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/gsap.min.js"></script>
<div id="root" data-composition-id="main"
data-start="0" data-duration="8" data-width="1080" data-height="1920">
<h1 id="title" class="clip" data-start="0" data-duration="4" data-track-index="0">Welcome to Acme</h1>
<p id="cta" class="clip" data-start="4" data-duration="4" data-track-index="0">Try it free</p>
</div>
<script>
const tl = gsap.timeline({ paused: true });
tl.fromTo("#title", { opacity: 0, y: 24 }, { opacity: 1, y: 0, duration: 0.6 }, 0);
tl.fromTo("#cta", { opacity: 0 }, { opacity: 1, duration: 0.4 }, 4);
window.__timelines["main"] = tl;
tl.seek(0);
</script>
- Timing lives in
data-startanddata-duration(seconds).data-track-indexlayers clips that overlap. - Size comes from
data-widthanddata-heighton the root; set the page'shtml, bodyto the same size in CSS. - Animation goes on the GSAP timeline, keyed to the same start times. Register it under the composition's id.
- Run
npx hyperframes lintafter editing; it catches missing attributes and timing mistakes.
Rendering from Code
import { createRenderJob, executeRenderJob } from "@hyperframes/producer";
const job = createRenderJob({ fps: 30, quality: "standard" });
await executeRenderJob(job, "./my-video", "./output.mp4");
createRenderJob takes the render settings (fps and quality are required; entryFile, format, variables, and others are optional). executeRenderJob takes the job, the project directory, and the output path.
Data-Driven Videos
For batch or personalized videos, generate one composition per row (a changelog entry, a customer, a metric), then render each:
- Keep one hand-built composition as the template.
- For each row, write a copy with the row's text, numbers, and images filled in, and the clip timings adjusted if the content length changes.
- Lint, then render each file with
npx hyperframes render -c <file> -o <name>.mp4orexecuteRenderJobin a loop.
Product announcements, changelog videos, testimonial cards, and stat reveals all fit this pattern: a short sequence of clips with text, an image or two, and one or two GSAP moves each.
Aspect Ratios
| Platform | Width | Height | Ratio |
|---|---|---|---|
| TikTok/Reels/Shorts | 1080 | 1920 | 9:16 |
| YouTube | 1920 | 1080 | 16:9 |
| Instagram Feed | 1080 | 1080 | 1:1 |
| Instagram Feed | 1080 | 1350 | 4:5 |
Hyperframes vs. Remotion
| Factor | Hyperframes | Remotion |
|---|---|---|
| Language | HTML/CSS/JS | React/TypeScript |
| Agent compatibility | Better (plain HTML) | Good (needs React knowledge) |
| Animation | GSAP timeline (plus CSS) | Spring physics, interpolation |
| Cloud rendering | Not built-in | Lambda (AWS) |
| License | Apache 2.0 (free) | Company license for commercial use |
| Ecosystem | New, growing | Mature, large community |
Use Hyperframes when: AI agent is generating the video, simple animations, batch templated content, cost-sensitive.
Use Remotion when: Complex animations needed, already using React, need Lambda for massive scale, want larger ecosystem.
Relevant Skills
- video
- social
- ad-creative