Collective Canvas
Generating a Real-Time Audio-Visual Collage from Spoken Prompts
Rather than sharing the thoughts and feelings of one artist, how could public art be used to reflect the thoughts and feelings of the community members who experience and interact with it?
Collective Canvas is an interactive installation prototype built in TouchDesigner that uses the Stable Diffusion API and other AI models to let people speak what they want to see and hear. The software seamlessly integrates each request into the generated artwork.
Collective Canvas Demo
AI allows for generating images of anything you can imagine. For this project, I was interested in creating a more natural way of engaging with the creative process with generative image models. Using diffusion, sentiment analysis, and regression analysis, this project allows a user to speak any content they wish to display and watch it appear. Ambient music also plays, matching the emotional quality of the provided prompt. The user can then play with a touch screen UI that interacts with a regression analysis model through Wekinator to further adjust audio and visual effects. The project concept could easily be translated to a larger-scale, public installation, enabling a community-designed artwork.