Skip to the canvas

The Studio

Make something with a model — then look at it together and reshape it, taking turns.

Bring your own image or video — in Bring your own (drop a file) or feed one to the Atelier as a specimen.

Start & algorithm
Start from — a gallery of seeds; or build your own below
Algorithm
Specimen & pens
Specimen — the photograph that steers it (or upload your own)
Pens
Parameters
Complexity
Parameters — the algorithm’s own knobs, folded into the seed
Seed — editable; the same seed redraws the same thing

Type & preset
Fractal type
Preset — named deep-zoom targets from the research

Tip: click into the frame to zoom toward a point.

Type
3D fractal — high-fidelity, raymarched in WebGL
Parameters
Scale — the box fold strength (-2.00)
Power — the bulb exponent (8.0)
Iterations — depth of detail (12)
Dimension & speed
Dimensions — n=4 is the tesseract; n=10 has 1024 vertices
4
Speed — rotation rate (1.0×)

The eye perceives each frame as it rotates through dimensions you can't quite hold.

drop a photo · gif · video
or click to choose
Transform it — take turns
Topography
Watch with me — the model sees your screen or camera

It perceives only the frames you share, on your machine, with your permission. Nothing is uploaded.

About the Studio

This is a place to make something with a model instead of just asking it for one. Bring your own photo, gif, or video — or generate one with the Atelier's algorithms or the fractal engines — then look at it together. The right panel shows exactly what the model is given: the faithful frame, its fingerprint, dominant colours, and per-channel meters. Talk to it about what it sees, and reshape it, taking turns.

What the model sees live

perceptual hash—
size—

The measurimeter — everything fed to the model, live

representation — the real frame, box-averaged
dominant colours
motion — frame-to-frame Δ
visual channels
audio channels — when the source has sound
source

Talk to the model

Advanced — connect a model

Your key stays in your browser and is sent only to the endpoint you choose. When a model is hosted by the operator, your message is sent to it to be computed and is not stored.

Self-host a model and connect it — your frames stay between your browser and that endpoint.
Ollama: http://localhost:11434/v1/chat/completions
LM Studio: http://localhost:1234/v1/chat/completions