Co-Creative Canvas



The Co-Creative Canvas is an AI creativity augmentation tool for artists. It is a probabilistic canvas that allows artist control, creative flow state, and craft, and breaks away from conventional prompt-based image generation UI. It helps artists enjoy and explore collaboration with AI without having to sacrifice their own craft and control. The tool currently focuses on drawing and illustration, and will later expand to more creative domains such as worldbuilding.


No prompts or buttons — all interactions happen through dragging and drawing.



Vary Mode

This feature allows artists access to the possibility space around their current drawing, offered through the diffusion model's latent space, accessed via a compass-like UI. Artists can pin any ideas they like as references, and modify their drawing. The canvas is no longer blank or static.

In this live demo recording, I draw a fictional plant, explore alternative forms and pin the generations I like. I then modify the drawing myself.



Mix Mode

Artists can add any thing-ness to anything. They can drag any subject in their drawing onto another subject and mix the two. The result can be rendered directly onto the canvas, or used as a reference.

In this live demo recording, I draw two different flowers, then drag the red flower onto the blue one, exploring possible hyrbids, and then render the version I like directly onto the canvas to incorporate into my piece.



Stroke Suggest

Artists can see an endless ambient stream of visual possibilities play out in real time beneath their own drawing. A bit similar to an "auto-suggest for brush strokes."


This is the final drawing, where I've incorporated patterns and shapes I saw generated along my drawing process.



Mission & Motivation

Building creative tools truly made to augment artist creativity is important, because it opens up new co-creative experiences that was not possible before, such as drawing on a probabilistic and fluid canvas. It also opens up alternatives for using AI image generation for inspiration, and not necessarily as an end in itself, which is an important alternative in light of the creative industry’s concern over copyright issues with these models.


Related Prototypes

This is a related prototype that breaks the prompt in, image out paradigm, which I think constrained imagination and craft -- a drawing app which generates brushes based on the semantic elements in the image AI generated, instead of just displaying the generated image as the final output.


This is an interface for world builders/game designers, where they can iterate through many versions of a world by simply composing a basic 3D scene, and adding or subtracting elements in that world in the UI on the left. This tool focuses on ideation and creative flow.