3 ms·

The "composite party photo", while impressive, shows that still the miniscule details are being lost, like the teeth structure of the guy in the middle or the f

The "composite party photo", while impressive, shows that still the miniscule details are being lost, like the teeth structure of the guy in the middle or the fact that the guy on the left is holding the cup with three fingers. Wondering why they chose this edit for the showcase.

nzach 27d ago
And don't forget the 'Thing' from Addams Family in the shoulder of the guy on the left.
- pllbnk 27d ago
  The guy on the left in that same photo only has three fingers (and a thumb, I suppose). I thought image generation has already outlived that. Edit: I feel stupid I didn't see the original OP already mentioning three finger issue. I'll just leave it here.
headz 27d ago
> Wondering why they chose this edit for the showcase. They probably didn't care to check those images in detail.
polytely 27d ago
I'm guessing if you work on these images the whole day you start to lose the ability to judge image quality. like your brain becomes slopified
vunderba 27d ago
The OpenAI PR/doc team seem to have a history of this kind of attention lapse - their 4-panel comic strip using gpt-image-2 was an absolute mess too. https://mordenstar.com/blog/gen-failures https://mordenstar.com/blog/gen-failures
- Caracas288 27d ago
  Another example was the doc for GPT-5, which had all its graphs messed up, but they said it was human error rather than AI.
kfarr 27d ago
I've also found the OpenAI image models to lose fine detail on image edits compared to Nano Banana or Flux models which faithfully retain input source image geometry and details. I was hoping this might be different but it sounds similar to previous OpenAI image models where something is lost in translation during image editing.
- vunderba 27d ago
  I know the API (assuming you're using it) lets you set the output size of the edited images - I haven't done a huge amount of testing with > 1mp, so I'd be curious if this might mitigate some of that. https://developers.openai.com/api/reference/python/resources/images/methods/edit https://developers.openai.com/api/reference/python/resources...
timbaboon 27d ago
Haha yes, I immediately saw the fingers and was surprised 'cos I thought that was something they had "solved" by now
- andai 27d ago
  Maybe it's like how spammers intentionally make their emails more obvious, because they have a very specific target audience.
- hmstx 27d ago
  Exactly this. "All this progress and they're back to miscounted fingers again?!"
nilsherzig 27d ago
Dogs shadow didn't change
fsloth 27d ago
The dog is even more on the nose. The shadow shape is almost the same despite the dog getting quite a lot of volume around it's original body parts.
nojs 27d ago
The guy’s arm is also extremely long.

nzach 27d ago

And don't forget the 'Thing' from Addams Family in the shoulder of the guy on the left.

pllbnk 27d ago
The guy on the left in that same photo only has three fingers (and a thumb, I suppose). I thought image generation has already outlived that. Edit: I feel stupid I didn't see the original OP already mentioning three finger issue. I'll just leave it here.

headz 27d ago

> Wondering why they chose this edit for the showcase. They probably didn't care to check those images in detail.

polytely 27d ago

I'm guessing if you work on these images the whole day you start to lose the ability to judge image quality. like your brain becomes slopified

vunderba 27d ago

The OpenAI PR/doc team seem to have a history of this kind of attention lapse - their 4-panel comic strip using gpt-image-2 was an absolute mess too. https://mordenstar.com/blog/gen-failures https://mordenstar.com/blog/gen-failures

Caracas288 27d ago
Another example was the doc for GPT-5, which had all its graphs messed up, but they said it was human error rather than AI.

kfarr 27d ago

I've also found the OpenAI image models to lose fine detail on image edits compared to Nano Banana or Flux models which faithfully retain input source image geometry and details. I was hoping this might be different but it sounds similar to previous OpenAI image models where something is lost in translation during image editing.

vunderba 27d ago
I know the API (assuming you're using it) lets you set the output size of the edited images - I haven't done a huge amount of testing with > 1mp, so I'd be curious if this might mitigate some of that. https://developers.openai.com/api/reference/python/resources/images/methods/edit https://developers.openai.com/api/reference/python/resources...

timbaboon 27d ago

Haha yes, I immediately saw the fingers and was surprised 'cos I thought that was something they had "solved" by now

andai 27d ago
Maybe it's like how spammers intentionally make their emails more obvious, because they have a very specific target audience.
hmstx 27d ago
Exactly this. "All this progress and they're back to miscounted fingers again?!"

nilsherzig 27d ago

Dogs shadow didn't change

fsloth 27d ago

The dog is even more on the nose. The shadow shape is almost the same despite the dog getting quite a lot of volume around it's original body parts.

nojs 27d ago

The guy’s arm is also extremely long.