In May 2025 I commissioned a team of motion designers to create a launch video. It cost us $2000 and 4-5 days of my full attention.

I noticed the video was likely built using a set of animations and transitions that this motion design team reused for all their customers. My theory was that if I could recreate that set of animation primitives and put them in a format coding agents could understand, I'd be able to create launch videos end-to-end way faster and have full control over the final product.

I spent 3 days setting up the repo and 4 hours building the motion video you see above. Those 3 days consisted of 1) building out the set of primitives and 2) building the iterative agent rendering loop so the agent could work autonomously.

The primitives I built were:

Text (word drag-in, flip-up, gradient fill, typing animation) Motion (object pop in, fade right, fade left, swoop, float, etc.) React components (logo, cursor, animated browser, animated terminal) Background designs (radial gradient, corner gradient, color fade, etc.)

The way I built these primitives was surprisingly simple. I leveraged Gemini's ability to analyze videos (Claude and GPT don't accept videos as a modality yet) by attaching the previous motion-designed video and giving it this prompt:

"Is it possible for you to thoroughly go throughout the video and, in detail, note all the different text transitions and what they look like, describing them in natural language so I can put them in my animation tool?"

It gave me a text file which I dumped into a coding agent (in my case Claude Code) to build animation primitives with. I repeated that for motion transitions and background designs, and modified existing React components to work with Remotion.

For building the iterative agent rendering loop, I gave the agent the ability to test its sections end to end.

Remotion can render individual frames via a bash command (npx remotion still {flags}) so I put in the AGENTS.md to heavily use that when designing sections. I also tried having the agent browser-use the Remotion editor spun up as a local dev server, but this was too confusing — it had to know how to scrub through frames in the UI, select the right sequence to check, etc. So I ended up sticking with the Bash command and found that coding agents could do a decent job testing animations by rendering frames 10-20 steps apart.

I'd love for you to try this out and welcome any feedback. Montage has really helped me create high-quality motion videos (an example is in the repo).

Comments URL: https://news.ycombinator.com/item?id=47263802

Points: 1

# Comments: 0

We gate full access to the platform partially for this reason. We debated giving fewer than 50 free browser sessions, for example, and have already banned a few accounts from our self-serve today that were unidentifiable/"company accounts" without a legitimate web presence.

We also have an example of a complex, multi-agent workflow here that might be useful for you to look at: https://www.simplex.sh/blog/context-engineering-for-web-agen...

Re: Puppeteer automation as part of the script -- we have a feature we wrote for one of our customers that we didn't promote to production where you can define a deterministic action in the dashboard that allows you to paste in JavaScript, but we're likely not to push that to prod anytime soon. Could you explain your reasoning for wanting to use Puppeteer still? We've generally seen customers fully switch over to Simplex instead of relying on their original Puppeteer/Playwright scripts -- since we have action caching, the underlying script (click on div locator with this div id, etc.) is pretty similar to what you'd get using Playwright.