---
title: Introduction
description: "Write HTML. Render video. Built for agents."
---
Hyperframes is an open-source framework that turns HTML into deterministic, frame-by-frame rendered video — so you can define a video the same way you build a web page.
## See It in Action
Here is a video defined entirely as HTML:
```html
```
Run `npx hyperframes render -o demo.mp4` and this produces an MP4 with deterministic, frame-by-frame capture. Same input, identical output, every time. No timeline editor. No proprietary format. Just HTML.
## Why Hyperframes?
**You already know the stack.** Compositions are HTML files with data attributes. Animations use GSAP, Lottie, CSS, or any runtime that can seek to a given frame. There is no custom DSL, no proprietary component system, and no React requirement. If you can build a web page, you can build a video.
**Agents already speak HTML.** Most video tools require complex APIs or drag-and-drop interfaces that agents cannot operate. Hyperframes compositions are plain HTML documents — the format LLMs are best at generating. An agent can compose, modify, and render videos using tools it already understands.
**Determinism by design.** The rendering pipeline is seek-driven with no wall-clock dependencies. `frame = floor(time * fps)` — every frame is independently captured via Chrome's `beginFrame` API and encoded with FFmpeg. Same input always produces identical output, making CI testing and batch rendering reliable.
Hyperframes was designed from the ground up for AI agent integration. Because compositions are plain HTML, any LLM can generate, edit, and iterate on video content without specialized tooling. Pair it with function-calling agents to build fully automated video pipelines.
## How It Works
Define your video as an HTML document. Each element gets data attributes for timing (`data-start`, `data-duration`) and layout (`data-track-index`). Add animations with GSAP, Lottie, CSS transitions, or any seekable runtime via the Frame Adapter pattern.
Run `npx hyperframes dev` to open a live preview at `localhost:3000`. Edit your HTML and see changes instantly — no build step, no compilation.
Run `npx hyperframes render -o output.mp4` to produce a final video. The engine seeks each frame in headless Chrome, captures it with `beginFrame`, and pipes the result through FFmpeg. Run locally or in Docker for fully reproducible output.
## Packages
Types, HTML parsing, runtime, and composition linter — the foundation everything else builds on.
Seekable page-to-video capture engine. Loads HTML in headless Chrome and captures frame-by-frame.
Full rendering pipeline combining capture and FFmpeg encoding into a single API call.
Visual composition editor UI for building and previewing timelines interactively.
Command-line tool for creating, previewing, and rendering compositions.
## Next Steps
Build and render your first video in 60 seconds
Understand the HTML-based data model behind every video
Add timeline-driven animations with GSAP
Render locally, in Docker, or in a CI pipeline