Project Description




AI technology combined with surrealist approaches. ‘Again’ is an investigation into our perception of repetition, creating landscaped environments through a combination of analogue and digital collage.  The exploitation of AI allows a focus on process and intent, with no fixed boundaries, the unified compositions shift in emphasis and remain in transit, neither dependent on lasting objects, nor fixed locations. The power of self-replication gives chaotic outputs, yet the context of each repetition informs the next. Through individual curation and guidance, the AI creates a surreal version of paradise for the viewer.


Created using pyTorch and an implmentation of “A Neural Alogorithm of Artistic Style”, Gatys, et al.

A Super Resolution GAN, ESRGAN, based upon “Real-ESRGAN”, Xinato Wang, et al. provided upscaling of final images to print scale resolution.

The generative process to produce the final pieces ran for ~16 hours on a high-end c. In addition, there were ~36 hours of computational time dedicated to model training, data production, upscaling, and video output. In total, over 1TB of data was produced.


Generative ai art installation//

42 inputs, 800 iterations and 124,800 total stills//

156 final artworks created//

The final display was a HD projection, displaying each image in it’s generative sequence, in an established psuedo-random order. Driven by a Raspberry Pi, the display can be controlled, sequenced, or randomised.//

Datasets for artistic stylisation and ai learning created during workshops exploring landscapes, through analogue and digital collages//

Individual images were combined in sequence to create over 4 hours of projection, showing the AI “as it works” in York Art Gallery. //