What is HyperFrames?
HyperFrames is HeyGen's open-source framework for writing a video as an HTML page. Clips, titles and audio are ordinary elements with data attributes for start time, duration and track; animation comes from GSAP, CSS, Lottie, Three.js, Anime.js, WAAPI or a custom adapter. The renderer seeks each frame in headless Chrome and encodes the result with FFmpeg, so the same input produces the same video.
It ships 7 agent skills, a core set of four plus Figma, slideshow and Remotion-migration skills on demand, and a catalogue of about 450 blocks such as transitions, captions, charts and maps. It installs as a Claude Code plugin with `claude plugin marketplace add heygen-com/hyperframes`, or through `npx skills` for Codex, Cursor, Gemini CLI and others. Renders run locally, in Docker, on your own AWS Lambda or Google Cloud Run, or on HeyGen's paid cloud.
HeyGen uses it in production, and the adopters list includes tldraw for pull-request walkthrough videos.
Is HyperFrames worth it?
HyperFrames is the strongest case yet for treating video as a web page, and it is moving at a pace that suits automation better than careful pinning. The repository has about 74,400 GitHub stars since it was created in March 2026; the npm package has 564 published versions, the latest v0.8.211 on 28 October 2026, and about 810,000 downloads in the week to 26 October.
The bet is that agents already write HTML well. A composition is an index.html file that plays in a browser with no build step, which is the main difference from Remotion, the project HyperFrames openly credits as its inspiration. Remotion asks for React components and a bundler and is source-available; HyperFrames is Apache 2.0 with no commercial-use threshold. Remotion's Lambda renderer is the more mature cloud path, which the HyperFrames README concedes.
Determinism has a condition attached. Local renders use your machine's Chrome and GPU, and the docs recommend Docker when a controlled Chrome, FFmpeg and font environment matters, so CI and production renders should run in the container. On AWS, a 480-frame render across 5 Lambda workers is shown costing an estimated $0.0214 before S3 transfer, and a render can fan out to 256 chunks.
Two defaults deserve a look before a team adopts it. The CLI collects telemetry that includes an environment fingerprint and the name of any coding agent driving it, links usage to your HeyGen email once you sign in, and opting out also removes you from canary rollouts. HeyGen cloud renders upload any media file of 5 MB or more as a HeyGen asset referenced by a public, unsigned URL. Neither is hidden, but neither belongs in a pipeline that handles unreleased footage without a decision.
The open issues read like a product in heavy use: Studio cannot open exports over 2 GiB, and a failed remote media fetch can be silently dropped from a render. For agent-written product videos, PR walkthroughs and data animations it has no open-source equal; for long-form editing it is still a renderer, not an editor.
Where it earns its place
A developer-tools startup shipping weekly releases
Turn each merged pull request into a short walkthrough video with animated diffs and captions
Install the Claude Code plugin and ask the agent to build the video from the diff, as tldraw does; render in Docker in CI so output is repeatable. The software is free; the cost is agent tokens and CI minutes.
A growth marketer at a SaaS company with 40 customer segments
Make a personalised product update video for every segment
Build one composition with template variables, upload it once to HeyGen's cloud and re-render with `--asset-id` and new values per segment. Paid in HeyGen credits, with 4K at 1.5 times the rate; per-credit prices are not in the docs.
A platform engineer running a video-generation service on AWS
Render many videos in parallel on infrastructure the company owns
Deploy the Lambda stack with three commands; renders fan out through Step Functions to up to 90% of unreserved concurrency. The docs' 480-frame example costs about $0.0214 in Lambda and Step Functions before S3 transfer.
Before you commit
- Setup
- `npx hyperframes init my-video`, then `preview` and `render`; needs Node.js 22+ and FFmpeg. Agent route: `claude plugin marketplace add heygen-com/hyperframes` then `claude plugin install hyperframes@hyperframes`, or `npx hyperframes skills update` for the core skills elsewhere. A desktop Studio app for Mac, Linux and Windows needs a free HeyGen account.
- Integrations
- Claude Code, Codex, Cursor, Gemini CLI, Copilot and other skill-aware agents; GSAP, CSS, Lottie, Three.js, Anime.js and WAAPI adapters; Figma import; AWS Lambda, Google Cloud Run and HeyGen cloud rendering; an embeddable player web component.
- Data and privacy
- Local renders keep everything on your machine. The CLI's telemetry sends command names, render performance and an environment fingerprint, and links usage to your HeyGen email after sign-in; `HYPERFRAMES_NO_TELEMETRY=1` or `DO_NOT_TRACK=1` turns it off. Cloud renders upload the project (200 MB archive limit) and serve media of 5 MB or more by public, unsigned URL.
- Cost at scale
- Free to self-run. On your own AWS, cost is Lambda GB-seconds at up to 10,240 MB per function plus a small Step Functions fee; the docs' 480-frame example is an estimated $0.0214. HeyGen cloud rendering is billed per credit, with 4K at 1.5 times; credit prices are not published in the HyperFrames docs.
- Limits
- 74 open issues, including Studio failing on exports over 2 GiB and long paths, and failed remote media fetches dropped silently; mixed audio is summed at unity with no limiter; skills via skills.sh can lag main by hours.
What's good
- Plain HTML compositions with no build step
- Apache 2.0 with no per-render or commercial fees
- Agent skills and a 450-block catalogue
- Local, Docker, Lambda, Cloud Run or HeyGen cloud renders
What's not
- Telemetry includes an environment fingerprint
- Cloud renders expose large media by public unsigned URL
- Releases several times a day; pin versions in CI
- Byte-identical output needs the Docker renderer
Best for
Developers and agent pipelines producing short product, data or documentation videos from code.
HyperFrames alternatives
HyperFrames: common questions
- Is HyperFrames worth it?
- HyperFrames makes video something a coding agent can write, and its Apache licence removes the bill that comes with Remotion. We rate it 4.4 out of 5.
- How much does HyperFrames cost?
- Free under Apache 2.0, with no per-render fees or commercial-use thresholds. HeyGen cloud rendering and HeyGen media (avatars, generated clips) are paid in HeyGen credits; 4K cloud renders cost 1.5 times as much.
- What are the downsides of HyperFrames?
- Telemetry includes an environment fingerprint; cloud renders expose large media by public unsigned URL; releases several times a day; pin versions in CI; byte-identical output needs the Docker renderer.
- Who is HyperFrames best for?
- Developers and agent pipelines producing short product, data or documentation videos from code.
- How much does HyperFrames cost for a team or in production?
- Free to self-run. On your own AWS, cost is Lambda GB-seconds at up to 10,240 MB per function plus a small Step Functions fee; the docs' 480-frame example is an estimated $0.0214. HeyGen cloud rendering is billed per credit, with 4K at 1.5 times; credit prices are not published in the HyperFrames docs.
- Is HyperFrames safe for company data?
- Local renders keep everything on your machine. The CLI's telemetry sends command names, render performance and an environment fingerprint, and links usage to your HeyGen email after sign-in; `HYPERFRAMES_NO_TELEMETRY=1` or `DO_NOT_TRACK=1` turns it off. Cloud renders upload the project (200 MB archive limit) and serve media of 5 MB or more by public, unsigned URL.
- What are the best HyperFrames alternatives?
- The closest alternatives we have reviewed are VideoAgent, OpenDesign and ComfyUI.
Answers checked Oct 28, 2026.
Make HyperFrames? Show this rating on your site
