I Turned Jordan Drake Into a Knight Using ChatGPT’s New Image Model

A split-screen image showing a man in casual clothing standing between two suits of armor on the left, and the same man wearing one of the suits of armor on the right.

ChatGPT has updated its image generator, which promises to be more faithful to source photos.

When someone uploads a photo into ChatGPT and asks it to make edits, what many people don’t realize is that the AI chatbot is using that photo in much the same way it uses text as a prompt.

This means that the resulting image created by ChatGPT is actually novel and is only drawing inspiration from the source photo rather than editing it. Last year, a police department in Maine got tripped up by this.

OpenAI, the company behind the popular chatbot, has attempted to address this issue in the newly released ChatGPT Images 2.5.

“The most meaningful images feel grounded in the real-life people, places, and memories they’re based on,” OpenAI says.

“Images 2.5 is better at working from reference photos to transform familiar subjects across new settings, visual styles, and compositions. Image subjects look more recognizable, lighting and textures feel more natural, and distinctive features are more likely to carry through.”

A hand holds a printed photograph of a woman in a magenta top against a green door, next to a digital version of the same image.
Example given by OpenAI.

ChatGPT has given a series of examples showing off its new, improved fidelity. But PetaPixel tried it out with photos of our very own Jordan Drake to see how it does.

First up, since I shot this photo of Jordan standing next to knights in armor, it made sense to make him a knight. I’ve got to say, the AI image doesn’t look very much like Jordan at all.

The second time around — and this is all the photo editing I’m allowed to do on my free account — it definitely did a much better job.

Granted, the Rolleiflex lettering is all mangled, the straps are see-through, and his fingers seem to be inside of the camera. But the AI image does unnervingly resemble PetaPixel’s beloved camera reviewer.

A man holding a camera stands in front of a stone archway connecting two sections of an old, ivy-covered stone building.
Can the real Jordan Drake please stand up?

Elsewhere, Images 2.5 can now turn doodles into images, and it now processes images a lot faster — 50% faster.

Overall, OpenAI says it can make images with “more natural lighting and richer textures,” and is better at following editing instructions “across multiple turns.”

ChatGPT users should be able to find Images 2.5 now on desktop, mobile, and web. It’s also available on Codex and ChatGPT Work.


Image creditsPhotographs by Matt Growcoot

Affiliate Disclosure PetaPixel articles may include affiliate links; we may earn a commission if you buy through one.

Discussion