How we rendered 20 to 30 videos a week

Our social videos were being made one at a time in Figma. The work was careful, but much of it was repeated: find the template, replace the products, update the copy, export the right size and check the brand details again.

We were producing about three videos a week. The format itself was already stable, so rebuilding it manually was no longer buying us much.

We treated the video like product UI

In Remotion, the repeated parts became components: product cards, type, transitions, captions and end frames. Brand colours, spacing and timing became shared values. The inputs that genuinely changed remained simple: products, images, copy, offer and duration.

This did more than speed up exports. A correction to caption placement or transition timing could be made once and carried into every new video.

More output changed what we could test

The system could render 20 to 30 variations in a week without asking someone to recreate each one. That made it reasonable to test several openings, product groups and calls to action instead of spending the whole budget on one polished guess.

Volume was not the goal by itself. It shortened the distance between an idea, a published version and evidence about whether the idea was worth keeping.

Automation moved the work

The system handled repeated assembly. A person still chose the audience, the hook and which output was fit to publish. We spent less time maintaining the format and more time deciding what the format should say.

That is the useful boundary for this kind of automation. Standardise the parts that should not drift. Keep judgement where the answer changes from one piece of content to the next.