Skip to main content
How to Make a Video with HyperFrames: A Step-by-Step Guide
Guides|June 19, 20264 min read

How to Make a Video with HyperFrames: A Step-by-Step Guide

The one-page, do-it-now guide: install, scaffold, drop in your footage and voiceover, describe the video to Claude Code or Codex, preview, and render an MP4 - with the reel we made this way.

Gabe KedingParker NewellLuke Keding

The OneWave Team

AI Consulting

This is the short version - a one-page, do-it-now guide to going from nothing to a finished MP4 with HyperFrames, the open-source "video as code" framework from HeyGen. For the why and the deeper background, read HyperFrames: Make Real Videos by Writing HTML. This page is just the steps.

Made by OneWave with HyperFrames + Claude CodeEverything below produced this - one HTML file, one render command.
We built this with the exact steps on this page: royalty-free and iPhone footage, a real founder voiceover, an HTML product demo, titles, transitions, and a music bed - all in one composition, rendered to MP4.

What you need first

  • Node.js 22+ and FFmpeg on your PATH (FFmpeg does the encoding).
  • An AI coding agent - Claude Code or the Codex app. It writes the composition for you.
  • Any raw material you want in the video: footage, a voiceover, a logo, screenshots. All optional - you can also do it purely with on-screen graphics.

Step 1 - Install the skills (once)

This teaches your agent the whole HyperFrames workflow. In a terminal:

  • npx skills add heygen-com/hyperframes (add --all for the full skill set: registry blocks, captions, media tools, animation adapters).

Step 2 - Scaffold the project

  • npx hyperframes init my-video - creates the project folder, an starter index.html, and copies in any media you point it at.
  • cd my-video and you are ready to author.

Step 3 - Drop in your raw material

  • Footage: put your clips in the project (AirDrop your iPhone video, or download royalty-free stock). They become <video> elements on the timeline.
  • Voiceover: record a take on your phone and drop it in, or generate narration with the built-in text-to-speech. Run npx hyperframes transcribe to get word-level timestamps for captions.
  • Brand: add a design.md with your colors and fonts and a fonts/ folder, so everything renders on-brand.

Step 4 - Describe the video to your agent

Open the project in Claude Code or Codex and just say what you want - for example: "A 30-second product launch video: open on the hero shot, three feature cards, a pricing beat, end on our logo. Use my brand colors and the voiceover I dropped in." The agent writes and edits the HTML composition - scenes, timing, animation, captions, transitions.

Step 5 - Preview as you go

  • npx hyperframes preview opens HyperFrames Studio with frame-accurate scrubbing and hot reload - every edit updates instantly.

Step 6 - Check it

  • npx hyperframes lint validates the structure and timing.
  • npx hyperframes inspect catches text spilling off-screen or overlapping.

Step 7 - Render

  • npx hyperframes render produces a standard MP4 (1920x1080 at 30fps by default). Flags let you change frame rate, quality, or export WebM with transparency.

Step 8 - Put it on your site

The output is an ordinary video file, so embed it like any other clip - host it and drop in a normal HTML <video> tag with a poster image. Because the source is code, it lives in version control: when your brand or copy changes, you re-render instead of re-shooting.

That is the whole loop

Install, scaffold, add material, describe, preview, check, render. We used exactly this process to build the sizzle reel on our main HyperFrames write-up - real footage, a founder voiceover, an HTML product demo, and our logo, all in one file. If you want us to wire HyperFrames into your team's workflow or build a batch of branded videos, book a free call.

HyperFrameshow to make videos with HyperFramesvideo as codeAI video generationClaude Code videoCodex videoHyperFrames tutorialOneWave AI
Share this article

Want this applied to your own stack?

Leave an email and we'll send back where this fits your team specifically — what to try first, and what it actually takes to run it.

No sequence, no spam. Or book a call directly.

Rather just talk? Book a free 30-minute call — no pitch, just a clear look at what's possible.

Not ready to talk? Stay in the loop.

OneWave Monthly: our best AI reads, blogs, and workflows. Unsubscribe anytime.