Toflr

Toflr investigates a different approach to generative media. Instead of asking an AI model to generate every frame, Toflr uses AI to create structured procedural representations that can be validated, modified and rendered locally.

This makes animation reproducible, programmable and independent from continuous cloud rendering.

The model generates the system

A frame-by-frame model has to redraw everything at every step, which is why consistency drifts and why the compute never stops. A procedural representation is a small, inspectable description of how the result is built. The AI writes that description; the device runs it.

• Consistency: a mathematical description does not drift between frames
• Control: the representation is parametric and can be edited by hand
• Reproducibility: the same description always yields the same result
• Local rendering: the output is produced on your machine

The device generates the result

Because the heavy step is executing code rather than sampling pixels, rendering happens locally and predictably. Nothing about a project needs to leave the device to become a finished animation.

Public support

The project has received public support from the Generalitat de Catalunya for its development.