HyperFrames is an open-source video generation framework developed by HeyGen that enables developers and AI agents to create videos using standard web technologies like HTML, CSS, and JavaScript. Instead of relying on traditional timeline editors or proprietary formats, HyperFrames treats video as a structured web document, allowing users to define scenes, timing, and layout through HTML attributes.

The core idea behind HyperFrames is “write HTML, render video.” Users define a composition using HTML elements annotated with timing properties such as start time, duration, and track index. These compositions can then be previewed in the browser and rendered into video files (e.g., MP4) using a CLI tool. The rendering process is deterministic and frame-accurate, meaning the same input always produces identical output—an important property for automation, testing, and reproducibility.

HyperFrames is designed specifically for AI agent workflows. Since large language models are already proficient at generating HTML, the framework allows agents to author complete video compositions without needing to learn specialized APIs or interfaces. It also includes a non-interactive CLI and a “skills” system that helps AI tools like Claude Code or Cursor generate valid compositions and control the rendering pipeline programmatically.

Under the hood, HyperFrames uses a seek-driven rendering engine that loads compositions in headless Chrome, captures frames individually using browser APIs, and encodes them into video using FFmpeg. This architecture ensures precise control over each frame and enables consistent results across environments, including local machines, Docker, and CI pipelines.

The framework supports integration with common animation tools such as GSAP, Lottie, and CSS transitions, allowing developers to create rich, animated videos without leaving the web ecosystem. It also includes a component catalog and optional studio interface for building and previewing compositions visually.

HyperFrames is positioned as an “agent-native” alternative to traditional video tools, focusing on programmability, reproducibility, and seamless integration into automated workflows rather than manual editing.

Key features include:

• Define videos using HTML with timing and layout attributes
• Deterministic, frame-by-frame rendering (same input → identical output)
• CLI for previewing, rendering, and managing video compositions
• Built for AI agents with prompt-driven composition workflows
• Integration with web animation tools (GSAP, Lottie, CSS)
• Headless Chrome rendering pipeline with FFmpeg encoding
• Supports automated pipelines, CI workflows, and batch rendering
• Open-source under Apache 2.0

Common use cases include:

• Generating videos programmatically
• Creating AI-generated video content
• Automating video production pipelines
• Building data-driven or templated videos
• Integrating video generation into developer workflows or CI systems

HyperFrames is developed by HeyGen, an AI video technology company, and released as an open-source project designed for agent-first development and automation workflows.