I was getting very high quality, optically realistic images with varying tones and natural lighting. Now the images come out super grainy, sterile and low quality no matter how much I try to force the prompts. Is this issue occurring for anyone else? Is this a backend tuning issue?
Welcome to the forum!
Can you be more specific?
Show prior positive results with current degraded results along with the prompts and configuration used, such as which model, settings, etc.
I recommend you to check this topic:Collection of GPT-image-generator 2.0 issues, bugs, and work-around tips (check first post)
There are others that are experiencing the same kind of issues and maybe share tips for a workaround, if someone discovers some.
I had a perfect pipeline last week it was - oh wait this is a month ago then no that doesnt fit but maybe, it was working amazing for me a week or so ago I was pumping out images in workflows I set up with astonishly good 10 image references that it used insanely well.
Now I have fought it for days trying prompt variations different APIs, many many many ideas. only now did I think to google if GPT Image 2 got worse, and it sound like many indeed are saying so… im curious if anyone else experienced this is same timeframe as me cause it is a major noticeable change - worse coherence to the reference images, more muddy and less clarity, it is very frustrating
Yeah… But not only GPT Image… NanoBanana also…
But image editing is still fine…
Sorry for the slow reply! Thanks for the detail about graininess and reference-image consistency. Could you share non-sensitive before/after examples with the same prompt, model and settings, and note whether you're using ChatGPT or the API? We haven't confirmed the cause or a fix. Current image editing options are here: https://help.openai.com/en/articles/11084440