Made an AI audio-reactive visual system trained on old album pictures of my childhood / What do you think?

After a deeply introspective and emotional journey, I fine-tuned SDXL using old family album pictures of my childhood [60], a delicate process that brought my younger self into dialogue with the present, an experience that turned out to be far more impactful than I had anticipated.

What's particularly interesting about the resulting visuals, is that they seem to be imbued with intricate emotions, and not-so-well-recalled distant memories. Intuition tells me there's something of value in these kinds of experiments.

First video is Archaia's TouchDesigner system intervened with the resulting LORA. And second one is a real-time test (StreamDiffusion) of said LORA + an updated version of Auratura working in parallel.

For the soundscape I've used mainly my Expressive Osmose + SOMA Cosmos combo.

You can explore more of my work, tutorials, and systems via: https://linktr.ee/uisato

Linktree doesn't exist

Sounds deep. Will have to have a proper delve into this. 

Some amazing talent with those videos. Have no idea what the process is. Perhaps you can give us the artist spiel?

Very cool! It's super fun to train and fine tune models on your own dataset. Especially if you're interested in the technical part as well, such a cool intersection between science, art and basically alchemy at the moment.