Towards the Recursive Mirror

After alienation and déjà vu, I feel like another thread is starting to form. Both already carry repetition inside them: alienation is like looking into a mirror that doesn’t return you fully, and déjà vu is memory looping back with no clear origin. I want to see what happens if I push repetition further, until it becomes the work itself.

Early Tests

I started running quick experiments. I take an image, feed it into the system, then loop the output back in again. With every pass, the picture drifts further away. First it’s still recognisable, then edges stretch, faces blur, and by the fifth loop it’s already something else.

It feels less like making images and more like watching memory distort. Each round keeps part of the original, but also loses something. The subject is still there, but not quite present anymore.

To do this I used a new vibe coding ability on Ai Google Studio called “Build” , I created a mini app that runs images with a simple prompt.

This service was introduced in the early 2025 and it's free to use, it helps users with zero knowledge of coding to create their app. This is my second try using the app. Previously I created an app that measured the emotion and based on that to generate an image of you. I might write about it later.

Prompt “A portrait processed recursively through AI, each new image generated from the previous output. With every iteration, the figure drifts further from the original, faces begin to fragment, colors distort, and the reflection slowly loses familiarity.”

For this I used the Imagen 4 model, The results are interesting. Towards the end it misses the subject entirely and comes up with nonsense images of other things. 

Interestingly it changes sometimes it changes the ethnicity or turns the image into black and white. 

Starting from the top going all the way to the bottom. 

Notes to Self

What I’m noticing is that recursion doesn’t only copy, it changes. The act of repeating creates distance. Familiarity turns into estrangement.

This is close to what I’ve been circling from the beginning: recognition collapsing into something uncanny. For now these are just test loops, but already they feel like they hold the metaphor.

The next step is clear: keep refining, and find a way to let viewers stand inside this tunnel of reflections.