Giving old Illustrator characters a second life in 3D.
Years ago I drew a small cast of flat, vector-style characters in Illustrator and Photoshop. One weekend I fed them into Nano Banana, Google’s Gemini image model, just to see what would happen. Could it keep the same pose, personality, and silhouette but render them soft and dimensional, without me redesigning anything?
- Type
- Personal Project
- Tool
- Nano Banana (Gemini)
- Status
- Weekend Experiment
Before & After
Same character, same pose, new dimension.
The goal wasn’t a redesign. I wanted to keep every illustration’s original pose, outfit, and expression intact while letting the render carry actual fur, fabric, and light. Drag each image to compare.
Fluffy — The flying cat
Fluffy has been floating through my sketchbooks for years. A round, sleepy tabby who drifts through the clouds like a stray balloon, more curious than worried about where he'll land.
Nano Banana didn't stop at a still render. It animated Fluffy drifting through the same clouds.
Nano Banana
Illustrator
Detail: fur, fabric, and light the original flat illustration never had.
Franky — The wolf on a bike
Franky is the neighborhood wolf everyone's actually glad to see. Scarf on, basket loaded with treats, always pedaling snacks over to share instead of chasing anyone down.
Nano Banana
Illustrator
Nano Banana also dropped Franky into a full scene. Same wolf, same bike, a park to ride through.
When they met
One last generation for fun: Fluffy and Franky, together in the same frame.
Nano Banana even put them in the same scene together. Fluffy drifting down to say hi while Franky waves from the bike.
The Approach
Reference-driven prompting, not text-to-image guessing.
Character identity first
Each original illustration went in as a direct reference image, not a text description, so the model had the actual silhouette, colors, and outfit to match rather than reinterpreting the character from scratch.
Same pose, new material
The prompts stayed focused on rendering treatment: fur, fabric, light, depth. I kept explicitly asking it to preserve the pose and composition, so Fluffy still floats through the exact same clouds and Franky still grips the same handlebars.
Where it needed a hand
Backgrounds, framing, and a few proportions still needed manual cropping and picking between generations. The model gets the character right far more reliably than it gets the full scene right on the first pass.
Closing note
A weekend experiment, not a redesign.
This was a self-contained exploration, not an ongoing series. Just a few old characters and an afternoon of prompting. If you want to talk through the workflow or try it on something of your own, I’m happy to chat.