@sketchmark/motion

Your AI agent can now produce video

The agent writes the animation. Sketchmark renders it. No GUI. No designer. No bottleneck.

01

Every agent pipeline hits the same wall

Text output? Done.
Code output? Done.
Images? Done.
Video? Dead end.

Every video tool assumes a human dragging layers on a timeline. There's no format an agent can reliably write and a renderer can reliably execute. Until now.

02

A video format agents can actually write

A structured DSL — not a GUI, not a prompt, not a black box.

circle dot r:20 fill:#60a5fa | 0s - 3s x: 100 → 500 over 2s ease-out opacity: 0 → 1 over 0.3s

That's a moving circle. An AI agent wrote that in one shot. No hallucinated APIs. Just valid, renderable output.

03

Why agents write it well

Deterministic

Same input, same output, every time

Low token cost

Describe what happens, not every frame

No invalid states

The schema is small and strict

Version-controllable

It's just text — lives in your repo

04

What your pipeline looks like

1User prompt
2Agent generates .mmark file
3MotionMark validates + renders
4MP4 output — ready to ship

No human in the loop. Structured input → finished video.

05

What agents are building

Product demo videos from a README
Algorithm visualizations from code
Pitch deck animations from bullet points
Data story clips from a CSV
Social content from a calendar

06

Open source. Yours to run.

Self-host it, embed it, extend it. No API key. No render credits. No vendor lock-in.

Install the package

npm install @sketchmark/motion

Install for your agent

npx skills add anmism/sketchmark-motion --yes

07

The AI Editor

Agents generate the first draft. You refine it.

Open the Editor