AI video creation
In March 2022, a series of new Generative Artificial Intelligence tools appeared that allowed us to carry out new graphic experiments: Vqgan+CLIP (with which I made this video) and Disco Diffusion were the ancestors of the tools that started after a few months to spread throughout the Internet. After less than 6 months, Midjourney arrived with a truly strong impact on the world of digital graphics.
Personally, I was fascinated by it and immediately began to deal with these new technologies for creating images and videos. Initially online, then, after equipping myself with a GPU that allowed me to perform complex calculations locally, on my PC, thus bypassing the increasingly stringent censorship.
In March 2023 I was invited by the Terzo Paesaggio cultural association of Chiaravalle Milanese to organize a 2-day event in which we could discuss how the man-machine relationship was changing and in the same period, on the occasion of the release of my new album of electronic music “Debris”, I was asked by the editorial staff of Frequencies.eu to write a reflection on generative AI.
It seems like a century has passed, but it was only 1 year ago… and today the tools have multiplied at the speed of light, while the expression “Artificial Intelligence” has become mainstream and on social networks it seems that one cannot avoid the talk about it, often inappropriately.
I think that as always this trend will pass, and as always this will simply be one of the evolutionary steps of digital arts.
In the meantime I started making AI Video, in particular creating fantastic and surreal worlds for video clips of musical bands who wanted to accompany their music with new and abstract imagery. Nothing of what I did was “done by AI” because in themselves these tools are not capable of doing anything: generating graphics, subjects and environments is undoubtedly a new form of ultra-technical skill and learning to write the right prompt to obtain a result is something that can only be refined after a long process of trial and error.
While on the one hand we have ready-to-use and very intuitive tools such as Midjourney (which requires a monthly subscription and is hyper-censored), on the other hand the opensource world of Stable Diffusion also continues to evolve. Working with a modular system such as ComfyUI, applying different models and waiting for the often very long generation and rendering times requires dedication and patience.
Examples of AI Video I realised for music bands:
Here after some links to watch some AI Video released in the past 2 years: