Stanley Virgint | infinitecanvas : Collaborative AI image generation canvas

Infinitecanvas is an experimental platform that empowers users to collaboratively create and expand images using textual descriptions, leveraging AI technology and community-oriented design to pioneer a new approach to online art and collaboration.

Infinitecanvas is an experimental platform that allows users to collaboratively create and expand an image using textual descriptions. Inspired by my research thesis, Expressive Data: Fostering Connections With Online Crowdsourcing Through Creative Applications, this project’s initial ideation explores the use of crowdsourcing in fostering connections and collaboration among participants in creative contexts.

The project intervenes in key debates surrounding online collaboration, methods of crowdsourced contributions, and empowering generative AI among participants in creative contexts. Mainly, Infinitecanvas aims to pioneer a new approach to collaborative online art by leveraging emerging AI technology and community-oriented design. Its unique contribution lies in enabling users to collaboratively grow an image through textual prompts, creating a shared creative experience that transcends geographical and cultural boundaries.

The technical setup of Infinitecanvas is centered around the integration of open-source machine learning models including stable diffusion, outpainting supports, and additional components such as cloud storage and databases. The project’s outpainting functionality and model configuration are based on a fork of lnyan/stablediffusion-infinity, a preexisting outpainting tool with a utility-centered purpose. The Infinitecanvas project “space” on Hugging Face powers the machine learning models on an Nvidia A10G GPU and is integrated into a custom website wrapper I developed with HTML, CSS, and JavaScript. The website wrapper is designed to guide users through the experience, providing intuitive controls and background information. The primary site is hosted on Heroku, while image data is stored on Firebase and updates with each visitor’s accepted contribution. A separate real-time database on Firebase continuously updates the project with the most recent image URL, keeping the canvas image in sync.

Infinitecanvas explores the creative, philosophical, and conceptual implications of a communal AI graffiti wall, encouraging a sense of creativity among its users as it forces them to consider the visual trail they want to contribute and its relation to the growing narrative. By allowing users to collaboratively grow an image through textual descriptions, the project creates a shared creative experience that challenges traditional notions of art and collaboration. The platform’s use of AI technology and community-oriented design not only democratizes the creative process but also encourages users to engage with one another regardless of geographical and cultural boundaries. This community-driven approach is a direct application of the research paper’s findings, which showcases the appeal of low-skill contributions (i.e. typing text to generate an image instead of hand drawing) and how this enables consistent and simple contributions from wide audiences of people. The project ultimately seeks to establish a new paradigm in collaborative online art, where users can participate in a collective creative experience driven by AI and human interaction.

Repository: https://github.com/stanleywalker1</div

 

Tags:#Collaboration#GenerativeAI