Any photo reimagined by AI?

Seeing is conceiving when it comes to ChatGPT Vision

Want to re-create photos with Dall-e3?

Here’s something I tried yesterday - again with ChatGPT’s new vision capabilities.

I started with this photo of our crazy fur family - incidentally named after the Schitt’s Creek kids - Dave (David), Lexi (Alexis) and Stevie (The Rosebud Motel’s receptionist, turned owner).

Uploaded image

My prompt was simply:

Please provide a detailed description of what’s in this photo

ChatGPT gave me a reasonable description; here it is:

Certainly.

The photograph showcases three pets: two dogs and one cat, positioned in a home setting. The closest dog appears to be of a curly-coated breed, potentially a poodle mix, and is black or dark grey in colour. Its eyes are prominently visible, indicating curiosity or attentiveness.

Behind the first dog is a second one, with a curly and tawny coat, possibly of a breed like the Labradoodle. This dog is looking directly at the camera, offering a clear view of its light-coloured eyes and moist nose.

To the right is a cat with a striped tabby pattern in its fur. The feline displays a mix of brown and grey tones, with striking greenish-yellow eyes, gazing calmly and directly forward.

In the backdrop, a well-lit living space is visible, featuring modern interior design elements. One can see a table, a rug, and a few other furnishings. The entire scene emanates a sense of domestic comfort and serenity.

So then I simply switched ChatGPT to Dall-e mode and pasted the description as my prompt so that it would generate images based on my description.

Here they are:

As you can see, images 1 and 3 are pretty on the money - the others curiously not so much - Stevie got turned into a cat.

This one still needs work

So, I thought I was pretty clever with this because I got a good result on the first go - and maybe it works well for pets!

I tried this prompt on both a photo of me and a photo of the house and got mixed results:

Uploaded image
Uploaded image

I can’t argue that the images do reflect the description that ChatGPT generated, but not really the original image.

So we won’t be doing Superheroes with Dall-e just yet, but you never know, maybe soon.

Happy Hunting 🚀

Join the conversation

or to participate.