EXPERIMENTS WITH AI IMAGERY’S POTENTIAL FOR BLIND USERS
Artificial Intelligence may help blind people use visual rhetoric in unexpected ways
STORY
Read about Concept Realizations’ first-of-its-kind AI imagery collaboration with blind researchers here.
Our colleagues wrote image prompts on a variety of subjects, including famous works of art, which we ran through Midjourney to create AI-generated images. Our project was half research, half experimental art project, dealing with perception, language, and visual artwork and visual semantics, from a very unusual perspective. The results illustrate some of the historic challenges of rendering language into imagery, and vice versa. They also demonstrate how AI image synthesis interacts with prior art, how image synthesis works, and its limitations.
The project hints at AI image synthesis’ strange potential for allowing blind people to communicate with visual rhetoric in unexpected ways.
PROJECT IMAGES
As of this writing, Midjourney is still in a public beta testing mode, and users’ galleries are only visible to Midjourney subscribers. To set up an account to interact with the project results and image parameters, see this tutorial. As soon as Midjourney releases its public gallery functionality, we will make the complete project viewable there.
If you already have a paid Midjourney subscription, our projects 4,000+ image results are available for inspection at its Midjourney gallery:
midjourney.com/app/users/1249b1eb-2cb3-400b-afb9-b5ed13ed52a5
Midjourney user: ConceptRealizations#9755
The images are sorted into collections for easy comparison between different parameter sets (Midjourney version 3, version 4a, 4b, etc.), command variants, and subject matter, such as animals, famous artwork, and nature, for example. The gallery allows visitors to copy and rerun the images’ text prompts and image parameters. To make the results more reproduceable, every image was generated from the same random seed value: 123456789.