HeadsUpAI

HeyGen Launches HyperFrames Templates for Parallel Video Rendering on AWS Lambda

HeyGen, an AI video generation platform specializing in realistic digital twins, launched Templates on Lambda for its open-source HyperFrames framework. The update introduces a way to define dynamic variables—like names and colors—directly in HTML compositions. This sits alongside the framework's video-as-code architecture to provide a native path for high-volume personalization.
Variable types
string, number, color, boolean, enum
Input size limit
256 KiB per render
Infrastructure
AWS Lambda and Step Functions
Batch concurrency
Configurable via max-concurrent flag
Access
Open-source CLI and TypeScript SDK

This shift transforms HyperFrames from a local tool into an industrial-scale production engine. By offloading rendering to serverless functions (cloud-based code execution that scales automatically), teams can bypass hardware bottlenecks. It targets the programmatic video market, offering a direct migration path for users of existing frameworks like Remotion.

You can now use the lambda render-batch command to dispatch thousands of personalized renders from a JSONL file. The system supports five variable types and handles asset resolution at render time to stay within cloud limits. The update is available via the HyperFrames CLI and a TypeScript SDK.

View the full update on hyperframes.heygen.com
HeyGen
HeyGen
@HeyGen
X

Your best HyperFrames composition is now a template swap the text, the colors, the clips render dozens of versions in parallel on AWS Lambda build once, scale forever setup docs in the thread ↓ https://t.co/Ige3lAd39v

9retweets43likes
View on X

Still wondering? A few quick answers below.

HyperFrames templates are video compositions built with HTML and CSS that accept dynamic variables. By adding a specific data attribute to the root HTML element, developers can define fields like text strings, colors, and asset URLs. These templates allow a single design to be rendered into many unique versions by swapping values at runtime.

Batch rendering uses a command line interface or SDK to dispatch multiple video projects to AWS Lambda simultaneously. The system uses a JSONL file where each line contains unique variables for a specific recipient. AWS Step Functions then orchestrates the parallel execution, allowing thousands of personalized videos to be rendered at scale without managing a dedicated server cluster.

HyperFrames is designed with a similar architecture to Remotion, specifically regarding how properties are passed to compositions. The framework includes a migration guide for developers moving from Remotion Lambda, as both systems use a comparable JSON-based variable injection pattern and share the same 256 kilobyte input limit for cloud-based rendering workflows on AWS infrastructure.

Variables are restricted to a total payload size of 256 kilobytes to comply with AWS Step Functions limits. Because of this constraint, developers should pass media assets like images or videos as URL references rather than inlining them as raw data. The rendering engine then fetches these external assets during the capture process to produce the final video file.

Developers can test templates on their own machines using the HyperFrames render command combined with a variables flag. This allows for instant previews of how different data payloads will look without deploying code to the cloud. A strict variables mode is also available to catch type mismatches or missing data before starting a large-scale production batch.

Share this update