I’ve been using DALL·E inside ChatGPT Plus (GPT-4o) for months, and I know what high-quality output looks like. But for the last 4+ days, image generation has taken a huge hit in quality — something is definitely off.
Renderings are way too fast resulting in blurry, distorted, low-effort images
Faces are often wrong, proportions are inconsistent, and details are completely botched.
Previously working prompts now produce rushed, mushy, painterly junk.
Cinematic prompts no longer get sharp, consistent results.
It’s like rendering time has been silently throttled, or the model is skipping its deeper visual passes. This isn’t just a one-off — it’s been happening for several days now.
Anyone else experiencing this?
Would love a response from OpenAI staff ,has something changed in DALL·E’s backend? Is this temporary? Is a fix on the way?
Thanks, I just want the great results we used to get back.