Artimuse
Exploring new interaction mode in human-AI co-creation
| Role | Product Manager, Initiator, Developer & Designer |
| Team | Yvonne Fang, Subhashini Ramkumar, Alice Nie |
Overview
This project encapsulates my thinking and approach towards designing AI-driven media ceation tools. I think the key to making such tools accessible to all might be to build an interface that abstracts away the complexities of prompt engineering and gently guides the user, which will promote flow state and encourage creative thinking. I am also experimenting with the idea of giving more agency to the AI creative “copilot”, and having them contribute new inspirations to the artist sometimes even without being prompted, inspired by this paper, which presents a helpful framework of human-AI interaction design.Artimuse is an AI image creation partner for game concept artists, environment artists, art directors, and people in similar fields to use in their ideation or conceptualization phase of their creative workflow. They can co-create images with Artimuse without any prior knowledge of prompt engineering. All they have to do is talk in natural language. Artimuse not only helps bring ideas to life, but also sees and understands images it creates with the user, so that it can provide more inspirations when user asks for them, or even proactively, thus simulating a creative partner and enabling a flow state for the creator.
Design and art creation in industry usually requires specialization in digital tools and artistic processes. It’s also time-consuming. With Artimuse, generation and editing is quick, trials and errors are enabled and encouraged, thus allowing free experimentation, all through natural conversation. With Artimuse, we tailor the AI to understand its human creative partner - what they say, what they create, and every vague creative idea they try to pinpoint but can’t quite get there alone.
A Colab Proof of Concept prototype can be found here.