Creator
$49a month
For a steady posting schedule.
- 6,000 credits / month (about 12 videos)
- Pro quality tier
- Up to 40 s, 1080p
- 2 videos at a time
- API access
- 90-day history
[Made for] Research labs
Research explainers with exact numbers.
Paper and result explainers, model and dataset release clips, conference teasers. The video is code, not a diffusion model, so the text and numbers on screen are the ones you wrote.
What you would make
Four kinds of video.
Paper explainers
The idea of the paper in three or four drawn steps, 15 to 40 seconds long.
Result clips
One chart, one message: bars grow, the line draws on, the key number is highlighted.
Release clips
A model or dataset release: the name and three facts, in a look that fits the lab.
Conference teasers
Title, session and room on screen, sized for a feed or a slide.
Example briefs
A sentence in, a video out.
The clips are style examples made by our team's agents, not the result of these exact briefs. Hover or tap one to play it.
-
Brief 1 · Data Story
Accuracy by model size: 71.2, 78.5, 83.9 and 86.4 percent for S, M, L and XL. Highlight XL. Title: "Does size still matter?"
Use this brief -
Brief 2 · Whiteboard
Explain attention in three drawings: a row of tokens, arrows between them, and one red arrow for the strongest link.
Use this brief -
Brief 3 · Hologram HUD
Our new dataset: a wireframe globe, a marker locking on to one region, readouts "2.1 M samples" and "40 languages".
Use this brief -
Brief 4 · Blueprint
The pipeline in four labelled parts: tokenizer, encoder, retriever, decoder. Assemble them at the end.
Use this brief
Why code
Text is a string. A bar is a number.
A diffusion model paints letters and digits as pixels, and often gets them wrong. Here an agent writes a program that draws each frame, so a title is a string and the height of a bar is a number taken from your prompt. The agent renders stills and checks them before the full render.
You should still read the result before you publish it. The agent can misread a brief, and it only knows the numbers you give it: it does not look anything up.
Styles that fit
- Data StoryNight-blue chart panel: bars grow, a line draws on, one highlight
- WhiteboardA marker draws the idea live: arrows, one red pen, an eraser sweep
- BlueprintCyan drafting sheet: parts plot line by line, dimensions, exploded
- Iso InfographicIsometric diorama explainers: flat three-tone solids, numbered pins
- Hologram HUDCyan light projection: wireframe globe, scan band, ticking readouts
- Swiss MotionStrict grid, heavy grotesk, red and black snapping into place
How it works
Three steps, two of them yours.
You
Pick a style
Each style is a tested look with fixed rules for palette, type and motion. See all of them.
You
Write the brief
One or two sentences. Put the exact words and numbers you want on screen in the prompt. Set length, aspect, captions and sound.
Agent
It builds the video
A coding agent writes the scene, renders stills, checks them and renders the full video. You download the mp4.
The plan that fits
Creator is the usual pick.
Creator includes the Pro tier, which gives the agent more review passes, and videos up to 40 s. There is no free tier. Compare all plans.
| Length | Standard | Pro |
|---|---|---|
| 5–10 s | 350 | 1,050 |
| 11–20 s | 500 | 1,500 |
| 21–30 s | 700 | 2,100 |
| 31–40 s | 900 | 2,700 |
Have a brief? Write it here.
The prompt goes with you into the app, where you pick a style and a plan.