loops
[2024]
Project Duration:
3 weeks
3 weeks
Collaborators:
Christian Noll
Jessica Shen
Christian Noll
Jessica Shen
Personal roles:
Programming
Programming
Inspired by William Basinski’s Disentegration Loops, this project sets up a feedback loop between two generative models. The output of one model matches the input of the other, in our case image-to-text and the reverse, being the types of the models used.
The pipeline used for generating the installation contents.
The images are used as the subject of an immersive experience, being generated in real-time. The APIs of Dall-E and Replicate’s Clip Interrogator are being called with each other’s output as input using a Python script, while a Max patch monitors for new images and produces dynamically generated audio and visuals when a generated image is added to the queue.
Surround layout of the installation. The images and audio manifest spatially in a cyclical format.
A new initiating image can be chosen from the audience at any point of the process, making the contents of the installation different in each run of the loop.
An example demonstration of the concept. The images are being generated in real-time through a Python script (bottom left) making the API calls, and loaded in Max (right) where the audiovisual effects are implemented.
Audience interaction through image upload in a single-screen layout
Singlle-screen layout - input image/expanding progress grid/latest generated image
Github Repository