Early stage · Seeking builders & believers

AI video that knows which character to move

Every AI video tool today starts with a single image and guesses which character should act. InkMotion lets you define each character separately — and command them independently.

See the demo Express interest

Current AI video tools were not built for characters

If you work with illustrated characters — for courses, comics, or brand content — existing tools make you fight the model every time.

01

One canvas, no identity

You upload a scene. The AI sees pixels, not characters. It doesn't know who is who.

02

Actions contaminate

Ask Character A to wave. Character B waves too — or the background warps. The model guesses.

03

No per-character control

There is no way to say "Character A does X while Character B does Y" in a single generation.

04

Illustrated styles break

Hand-drawn and illustrated characters suffer the most. Realism-tuned models reinterpret them every time.

The approach

Define characters by image. Direct them by scene.

InkMotion's core idea is simple: each character is an uploaded reference image, not a text description. The system matches those references to the scene, then animates each one per your instructions.

STEP 01

Upload your scene

Your pre-composed first frame — background, characters positioned exactly where you want them.

STEP 02

Upload character references

Separate reference images for Character A and Character B. These define identity — not position, not description.

STEP 03

Give each character instructions

Per-scene action prompts for each character independently. "A walks left. B turns and points."

STEP 04

Generate

The system detects character locations, builds precise prompts, and generates your video.


Demos

See where it is — and where it's going

These are honest demos. The current build is a proof of concept. What matters is that the architecture works and the direction is clear.

Demo video coming soon

Proof of concept — per-character direction

The core workflow: scene upload, character references, independent action prompts, generated output.

Concept
Demo video coming soon

Current limitations — what existing tools do

Unguided AI video generation: action contamination, character confusion, style drift.

Limitation

This is early. The right people will shape it.

Whether you're a potential user, an engineer who wants to build this, or an investor who sees the opportunity — leave your details and we'll be in touch.

No spam. Ever.

Thanks — we'll be in touch. You're early, which means you matter.