Open Design Motion: What It Is, How To Set It Up, And Which Models To Run It On

Open Design Motion is the free, open-source motion design feature of Open Design, and it turns one written prompt into an editable motion video and an MP4.

It shipped in Open Design v0.25.0 on 15 October 2026, a release codenamed "OpenDesign Motion" and billed as "The Open-Source Alternative to Claude Motion".

The code is on GitHub at nexu-io/open-design under the Apache-2.0 licence, so the tool itself costs nothing.

I'm Julian Goldie, and in the video above I plug it into my Agent OS and build real motion designs with it.

In this post I'll explain what it is, how to set it up and how every feature works, and then I'll show you how the free models that can drive it score on GoldieBench.

If you want my team to look at how content like this can win you Google and AI search traffic, you can book a free SEO strategy session.

What is Open Design Motion?

Open Design describes itself as "the 1st collaborative AI design agent workspace", and it is local-first and open source.

Motion is the part of that workspace that makes videos and animations.

Your agent writes the animation as HTML, CSS and GSAP code, and HyperFrames renders the code into an MP4 through headless Chrome and FFmpeg.

HyperFrames is HeyGen's open-source, agent-native video framework, and Open Design builds it in as a first-class feature.

How to install Open Design Motion

The README recommends the desktop app as the fastest way in, with no Node, no pnpm and no clone needed.

macOS on Apple Silicon or Intel and Windows x64 get native apps from open-design.ai or GitHub Releases.

Linux has no prebuilt app in the official releases yet, so the README says to run it from source.

git clone https://github.com/nexu-io/open-design.git
cd open-design
corepack enable && pnpm install
pnpm tools-dev run web

The README says to open the URL that tools-dev prints, and it lists Node ~24 and pnpm 10.33.x.

To use it inside a coding agent with no UI, the README's one-liner is od mcp install <agent>.

On a Mac, the README warns that od may resolve to Apple's built-in octal-dump tool, so copy the snippet from Settings → MCP server instead.

How to use Open Design Motion, feature by feature

These are the features I showed in my own test, plus what the official release notes add.

My real test with Open Design Motion

My first prompt asked for a 20-second paper cutout story that worked as a trailer for Open Design.

It built the full motion design in the background while I did other work.

My second prompt asked for a 20-second high-energy reel titled "The AI Profit Boardroom community is the best to learn AI automation".

From what I saw, the results were genuinely better than what my normal human team can do, and that is my opinion from my own tests.

I then post the videos automatically with Blotato, and its analytics show me which platforms perform best.

If you'd like help turning videos and pages like these into rankings, grab a free one-to-one strategy session and we'll look at your site, your competitors and the keywords you're closest to winning.

How to write a good Open Design Motion prompt

A clear prompt saves you more time than any setting in the app.

First, say how long the video should be, such as 20 seconds.

Second, name the style, such as paper cutout, high energy or a clean product promo.

Third, give the exact title or line of text that must appear on screen.

Fourth, say what the video is for, such as a brand trailer or an Instagram reel.

Fifth, set the BPM and the style options before you send it.

If the first render is close but not right, ask for one change at a time rather than rewriting the whole prompt.

Because the video is code, a small change like a new colour or a slower scene is quick for the agent to make.

A simple weekly workflow with Open Design Motion

This is the loop I'd give anyone on their first week.

After a few weeks, Blotato's analytics tell you which platform deserves more of your videos.

Open Design Motion vs Claude Motion

The idea is similar, because you describe a video and AI builds it.

I prefer Open Design, because it looks more interesting and clean to me and it is easier to organise all my motion designs in one place.

It is also open source, so you choose the agent, the model and where your projects live.

Limits to know first

Linux users have no prebuilt app yet and need to run it from source.

The od name clash on Macs is a small trap, and the Settings → MCP server snippet avoids it.

There is an optional paid Design Plan with hosted models, and you don't need it if your own agents work.

The README says scrubbed safety and reliability telemetry is always on, while product analytics and session replay are consent-gated.

Output quality depends on the model you connect, which is where GoldieBench comes in.

Does Open Design Motion ship its own model?

No, it doesn't.

The README says Open Design doesn't ship an agent, and the agents already on your machine are the design engine.

You don't need Opus 5.5 to use it.

Inside OpenCode I showed free models such as LongCat, MiMo v2.6, Nemotron 3.5 Super, GLM-5.3 Flash and DeepSeek V4.1 Flash, which were free in OpenCode when I recorded.

I often use those same free models as the brain for Hermes.

So the useful question is which of those brains builds the best one-shot visuals.

How the free Open Design Motion brains score on GoldieBench

GoldieBench gives every model the same one-shot prompt to build a game, a page, a simulation or a visual, then scores the rendered result from 0 to 10.

That is close to what Open Design Motion asks a model to do, because both jobs are one-shot HTML that has to look good when it renders.

Here is where the brains I named stand on the live board, plus the nearest benched versions where the exact model isn't there.

For comparison, the paid model I said you don't need, Claude Opus 5.5, averages 7.57.

On those numbers, LongCat-2.0 is the free brain I'd try first, with the caveat that its score comes from only 4 tasks.

GLM-5.2 is the steadier bet with 47 scored tasks behind it.

Which models are best at visual and motion-style work?

The closest category on the board is Visual, where tasks like aurora, fireworks, lava lamp and synthwave test whether a model ships a polished, animated visual first time.

The live category page names Claude Opus 5 as the top model in that category, and Opus 5 averages 8.27 overall.

The same page notes that GLM-5.2 tends to dominate the Visual tasks, which makes it a strong pick if you want an open-weights brain for motion work.

If you want to run a brain on your own machine, check the local models board before you pick.

My practical advice is to start Open Design Motion on a free model, then switch to a stronger one only for the videos that really matter.

Who should use Open Design Motion

Creators who need lots of short videos should install it today.

Small businesses should use it for trailers, promos and explainers.

Agencies should test it for branded client videos and shareable live dashboards.

Anyone who already runs free models in OpenCode or Hermes can plug those straight in.

If you want my team to show you how to turn this kind of content into Google and AI search traffic, book a free SEO strategy session.

If you also want those videos found in AI answers, my post on GEO examples shows how I get named in Google's AI Overviews.

Also On Our Network

• Agent Operatorsthe seven-step Open Design Motion setup walkthrough
• AI Income Deskseven ways a business can use Open Design Motion
• AI Tool Verdictevery Open Design Motion feature, scored with a verdict
• agentos.guidethe Open Design Motion quick start for an Agent OS
• agentos.guidemy HyperFrames Keyframes guide on agentos.guide

If you want a clear plan for getting your content found in Google and AI search, book your free strategy session with my team.

Pick a free brain from the live board, give it one clear prompt, and you'll get the most out of Open Design Motion.