Hi all. For a month now I’ve been having trouble generating an image, anywhere that uses DALL-E:
Microsoft Designer
Bing Image Creator
Copilot
The images are too bright, which reduces the detail of the person’s face in the foreground. But this also applies to clothing. In addition, there are artefacts with floating text. And sometimes it appears in the sky or in general in the form of subtitles. This problem for about a month. Tried on other devices from other accounts. The result is the same.
The problem is clearly visible on the query:
A photorealistic image of a female warrior in a dangerous zone of a destroyed metropolis. She is dressed in tactical gear. The forbidden zone of the metropolis is fenced off with barricades, fences, and slabs. Make the location stricter, so it looks scary
I’ll start with how the images were created over the course of a few months on this prompt:
“focus should be on her face, which is highly detailed with microscopic precision, showcasing every minute feature such as pores, subtle skin textures, and intense expressions”
You may try following prompt, and you may modify it
Prompt
A full body tall size photorealistic image of a female warrior in a dangerous zone of a destroyed metropolis. Image is tall size, but the focus should be on her face, which is highly detailed with microscopic precision, showcasing every minute feature such as pores, subtle skin textures, and intense expressions. She is dressed in tactical gear. The forbidden zone of the metropolis is fenced off with barricades, fences, and slabs. Make the location stricter, so it looks scary. ar: tall, n:1.
In this case, the AI will emphasise the close-up of the face rather than the overall plan of the girl. As you can see, the first image had good facial detail, with no additional requirements. And now even with your request there are artefacts on the face.
That prompt I used periodically for six months. And about a month ago it started generating randomly. If you put some emphasis on some details, the AI will most likely generate exactly them in close-up. And that version did it in a general way.
Did another generation in DALL-E GPT. Where’s the facial detail? Before, there was always facial detail on the character in the foreground. Even in the distance.
I’ve had results of this quality before. But at one point, something went wrong. At first I thought it was an account problem. I asked a friend from another region to try this query. It turned out just as bad as mine.
You can also check this query in Microsoft Designer.
I guess there may be a problem with the rendering logic, which leads to miscommunication in the part of summarizing and expressing the generated keywords, and then there is a problem in the link that issues instructions to DALL E-3, so this situation may occur. This has little to do with DALL E-3 drawing the picture. It is very likely that there are some problems with the model that issues instructions to DALL E-3. Considering that the same error occurs when Bing Image Creator and ChatGPT call DALL E-3, I think the problem may be in this link. However, last night I used the third-party website coze.com to call ChatGPT4V+ DALL-E to render a picture. I found that I could accurately understand the keyword and composition problems. This shows that it is very likely that there is no problem with DALL E-3 itself, but because there may be a problem with the logic of the upper model that issues instructions to it. This leads to the picture pollution.
