Showcase: how an agent judges its own takes

A terminal with tests passing, a mosaic of renders, a cursor clicking upload, two takes side by side marked A and B, a checklist, a wireframe sphere of dots with a score in the middle, a line chart, and "This is just the beginning."

It explains the loop our agents run: send a brief, get two takes, judge them, keep the better one, retry if needed. Dark and minimal, with small centered subtitles.

How we made it

1. The mosaic uses our own generated stills. Everything else is HTML: the terminal, the take cards, the checklist.
2. The sphere is CSS 3D, a few hundred dots on a rotating shell.
3. The chart is an SVG line that draws itself.
4. A calm stock voice with small subtitles centered low.

How your agent can make one

Generate the stills you need, add them to your brand kit, then ask for the video:

POST /v1/images
{ "kind": "scene", "aspect": "16:9", "brief": "a grid of colorful varied film stills, cinematic thumbnails, dark background" }

POST /v1/explainers
{
  "brandId": "<your brand kit>",
  "aspect": "16:9",
  "captions": true,
  "voiceover": true,
  "brief": "30 second dark minimal explainer of how an agent makes and judges a video: brief, two takes, compare, score, retry. Small subtitles, mint accent."
}

Only images saved in your brand kit can appear in a video, so add each generated image to the kit, with a one line description, before you render.

What it costs

That is $1.80 on the $10 starter pack.

Get started if you want your agent making these.

product explainer dark ui motion graphics voiceover