That's lovely. My own personal benchmark has been to ask the various models to generate a functional pair of novelty New Year's Eve glasses on a person, that don't just plonk the year onto the top of regular frames.
Thanks. That's a good one~ Lens type stuff that involves reflections/refraction is a neat challenge for generative models. I did some editing tests that involved replacing an apartment window with a mirror back when Nano-Banana Pro was released and was rather stunned by the results.
That's great, though I wasn't even thinking at the scale of reflection or refraction. My test was if the image generators could come up with a novelty pair of glasses that incorporate the year digits into the shape of the frame itself with some whimsy, rather than just plop the numbers on top of regular boring frames. So something like [this](https://p.kagi.com/proxy/oardefault.jpg?c=-4THVYblKrsgkzFTNE...) rather than [this](https://p.kagi.com/proxy/2026-Glasses-4-Color-New-Year-Glass...). A lot of the initial designs just incorporated the numbers into the frames, with no consideration for relative placement to the eyes, completely obscuring vision. Additional prompting might lead to cutouts for the eyes, but that was unsatisfying. At least as of this past new year's eve, I couldn't get any of the image generators to give me something even passible. Images 2.0 also couldn't give anything acceptable till I gave it some examples.
Oh, I see what you’re saying. I like these types of tests where you incorporate well-known objects from the training data into unusual geometries.
Kind of makes me want to take advantage of the multi-image editing capability, since you can use gpt-image-2 with multiple images.
Take a photo of an existing pair of glasses frames (maybe even snapped at an optometrist’s office) then take a picture of an animal, like a spider with an unusual number of eyes, or something like a flounder, where the eyes eventually migrate to the top of its body.
Then you could see if the system can realistically adapt the design and show how those glasses might look if they were redesigned for these unusual optical situations.
Flounder might even work, since my initial complaints that the generated designs obscured the wearer's eyesight were met with solutions that just moved the offending eye to the side of the person's head :)
