gpt-image-2

“the only way you get image models that basically do reasoning is you glom a diffusion head on an LLM backbone, which is basically gpt image 2.0 and above”

“just generated with the built in image gen, it does have an advantage there over Claude”

“GPT-Image-2 already scored unsurprisingly very high: 12 out of 15 on text-to-image”

“best image generator that I have tried, however it has one massive disadvantage: draconian censorship”

“far easier to steer and is significantly better in terms of prompt adherence”

“GPT 2 is almost always better (80+% of the time?) than Nano Banana 2, albeit slower”

“gpt-2-image has such a yellow tinge though. It stands out badly on nearly any screen.”

“the only one that consistently works for me and renders what I describe in the way I describe it”

“This was true until late April, when OpenAI's GPT Image 2 overtook Nano Banana 2.”

“the image generation parameters of GPT-image-2 are largely ineffective”

“good luck getting them to generate sprite sheets with consistent frames for animation”

“It just breaks down in subsequent editing (around 3-4 editing prompts).”

“the censorship system is really horrible if any two humans touching each other in sport is now crossing the line”

“gpt-image series is still affected by the yellow saturation issues”

“You still have to do some manual work if you want to create something like a comic”

“It's crazy what the latest version can pull off, with reference images, text, etc.”

“It was (and still is) amazing to me that GPT Image 2.0 was this good at making a coherent manga page in one shot.”

“interesting that GPT Image-2 managed to 2-shot this with thinking turned on”

“Now we just need open weights Nano Banana Pro / GPT Image 2, and Seedance 2.0 equivalents.”

“it's crazy how much things have improved on the image model for OpenAI”

“it's finally consistent enough to use to generate app store screenshots that look incredible”

“my initial complaints that the generated designs obscured the wearer's eyesight”

“I believe it does a way better job than the other image generation models”

“Reasoning before rendering is the thing that separates it from the pack rn”

“gpt-image-2 is VERY flexible with the resolution. You can use arbitrary resolution within the max pixel budget.”

“OpenAI actually has really good adherence, but occasionally tends to introduce its own almost equivalent of "tone mapping"”

“Octopus: good try, almost a pass, but socks don't cover the ends of all the tentacles”

“There are a couple of AI-esque misspellings, but yeah the quality is remarkable, and rather scary.”

“It's still relatively trivial to create a moderately complex prompt and have it fail miserably.”

“I was quite impressed with how faithful it adhered to the prompts given”

“gpt-2-image has surprisingly been doing much better on more objective criteria in my test cases”

“you can definitely generate true transparent PNG files through the ChatGPT interface”

“Images 2.0 als produces a lot of low frequency artifacts they are in all directions”

“The quality of the text is really impressive and I can't seem to see any artefacts at all.”

“The fidelity to text, the photorealism, the ability to show the same character in a variety of situations”

“I'm been impressed when testing this model today, but it still can't consistently adhere to the following prompt”

“this looks like parity with human excellence, there isn't much space for further improvement”

“the best images from this model are so visually flawless that the only way to tell they're fake”

“doesn't seem to have any "slop" standard look, the images it produces are quite diverse”

“242 Elo points clear of the next best model and 93% win rate against random models”