Turning real stories into an AI-illustrated exhibition

Problem

A non-profit wanted to raise awareness of people’s real, difficult experiences — and to do it in a way that genuinely stops people and makes them care, both online and at in-person exhibitions.

Approach

I built a generative-AI platform with a strict rule at its core: the model was never allowed to invent. Someone shares their experience, and the system:

• rewrites it into a full, readable short story — vivid, but staying faithful to the account and adding no new facts (deliberate hallucination control);
• generates an image to accompany each story;
• packages story + image for both the website and a printed exhibition.

At the exhibition, the striking images draw people in; on the back, the story and a QR code let them go deeper into the issue behind the picture.

Result

My first project using AI for good — and the engagement bore it out: the site averaged about 2 minutes per visit, the highest I’d seen on that analytics panel for anything that wasn’t a purchase flow. People didn’t glance and leave; they stayed and read.

Notes

The genuinely interesting engineering problem here is one that matters everywhere now: making a generative model faithful — expressive, but strictly grounded in its source, inventing nothing.