Super Super Super Resolution

This generated video explores what happens when we use super resolution models to enhance our footage, either when restoring old films from low resolution recordings, or when trying to augment a photo taken on a small camera sensor.

In a playful manner, we repeat it over many iterations, thus highlighting the small changes this translation adds to the original image. Undeniably we are mixing our original photograph's content with noise and low-level detail information learned from other (sometimes proprietary) datasets.

We use the SFTGAN model which, when repeated, alters the original imagery with an effect like that of an analog film burning, when a frame gets stuck in the projector. This effect has been used in experimental cinema (for example in Bardo Follies (1967) by Owen Land) to visually represent film death on the screen, revealing the materiality of the film itself. In this way the audience can experience death and then still go home unharmed.

Recorded reality is more and more encoded and mixed with the artefacts of the neural age. This offers possibilities for creative expression, as long as these enhancements are in our hands. We ask alongside with "The end" screens: imagine the possible futures, what happens next?

This video has been shown in person at the Czech Design Week of 2021 and then online at the Machine Learning for Creativity and Design Workshop at NeurIPS 2021.

• https://neuripscreativityworkshop.github...
• https://github.com/previtus/SuperSuperSu...
• https://youtu.be/MtWk2GoHX9A?t=242