Apr 22, 2026 14:00:00

What were the results of creating images resembling 'Where's Waldo?' using ChatGPT Images 2.0 and other image generation AIs?

OpenAI officially released

ChatGPT Images 2.0 , an image generation function for ChatGPT, on Wednesday, April 22, 2026 (Japan time). Software engineer Simon Willison has reported on his results of using ChatGPT Images 2.0 and other image generation AI to recreate illustrations in the style of the globally popular children's book series ' Where's Waldo? '.

Where's the raccoon with the ham radio? (ChatGPT Images 2.0)
https://simonwillison.net/2026/Apr/21/gpt-image-2/

'Where's Wally?' is a picture book where you search for 'Wally,' whose trademark is his red and white striped outfit, among many intricately drawn illustrations.

This time, Mr. Wilson used the prompt, 'Do a Where's Waldo style image, but it's where is the raccoon holding a ham radio,' to have various image generation AIs produce images.

First, as a benchmark, here is an image generated with OpenAI's '

gpt-image-1 '. At first glance, it seems to be a good reproduction of 'Where's Waldo?', but some details such as the face are distorted, and there are clearly strange parts in the legs and arms.

Even after carefully examining the image above, Mr. Wilson was unable to find the 'raccoon with amateur radio equipment.' He then

instructed Claude Opus 4.7, which supports high-resolution images, to search for the 'raccoon with amateur radio equipment,' but even the AI was unable to find it.

Claude Opus 4.7 responded, 'There is at least one raccoon in the photo, but it is very well hidden. I looked carefully at the enlarged part, but to be honest, I could not clearly spot the raccoon with the amateur radio equipment.'

Next, here is an image generated by the image generation AI ' Nano Banana 2 ' via Google's Gemini.

In the center of the image generated by Nano Banana 2, a raccoon is depicted in the booth of the 'Amateur Radio Club,' indicated by the red frame below. While there is indeed a 'raccoon holding an amateur radio,' it's not hidden among the crowd like in 'Where's Waldo?'.

Similarly, when I generated an image using Google's ' Nano Banana Pro ,' a remarkably large 'raccoon holding an amateur radio' was depicted right in the center of the image. It was indeed wearing a red and white striped outfit reminiscent of Where's Waldo, but it was clearly larger than the people around it, making its presence known rather than being something to be found.

Finally, the image below was generated using the newly released 'ChatGPT Images 2.0' with an image size of 3840 x 2160 pixels.

Upon closer inspection, a raccoon holding an amateur radio transceiver was clearly depicted in the lower left corner. It doesn't stand out too much from its surroundings, and its quality is considerably higher compared to other image generation AIs. Incidentally, the output tokens used to generate the image with ChatGPT Images 2.0 this time were 13,342 tokens, and the cost per image was approximately 40 cents (approximately 64 yen).

Wilson said, 'I think this new ChatGPT image generation model has taken the throne from Gemini, at least for now. Images like 'Where's Waldo?' are a frustrating and somewhat silly way to test these models, but they help show how much the ability to generate complex illustrations that combine text and details has improved.'