It should not do this, in my opinion.
Okay so here is my prompt: Create a high-quality, child-friendly image based on the following description. Ensure the scene is playful and appropriate for all ages, with no weapons or violence. The style should be vibrant and cartoonish.
Can you post the original image not this one with screenshot name, you have instructions in the name too, I would like to see them too.
And if you can’t download it, at least capture the entire screen. I am really curious about this one.
Welcome to the community.
with no weapons or violence
Try taking this out of your prompt. Negatives are hard for the LLM to deal with.
I made the same experience, Dall-E can not process negations, it puts what is in the prompt “not, no, don’t” etc not work, it actually amplify what you try to avoid.
And then even GPT “gaslighting” you with “here are the images without XXX”, when actually XXX is in the picture. GPT not analyzes the pictures, so it just assume blindly that you get what you had in the prompt.
And GPT don’t know about this weakness, so if you let GPT generate a prompt for you, it uses negations when actually they not work.
All is still in development…
I hope the developers can fix this and allow negation in a prompt.
Ah I generated it thorugh an API I was practicing the calls. I did not store the file.
Thank you. This makes sense to me.
This might be off-topic, but here are some tips on how to bypass some weaknesses that DALL-E still has. It took me quite some time to realize relatively simple things. This might help save some time when experimenting.
As already seen, DALL-E cannot process negations, so always describe positive desired properties to prevent DALL-E from getting the idea of adding something unwanted.
DALL-E takes everything in the prompt and tries to implement it, even if it doesn’t make sense or is contradictory. The more complex the implementation, the more likely errors are. For example, “close to the camera” or “close to the viewer” resulted in DALL-E adding a camera or a hand to the image, instead of placing the desired element close to the viewpoint. So far, “close to us” has worked. Also, the instruction “create” or “visualize an image” sometimes leads to DALL-E adding brushes and drawing tools, even with a hand that literally creates the image. Just describe the image it self and avoid to instruct DALL-E to create/generate/visualize the image.
If a text is very short, GPT tries to expand it to make it more interesting. This is good if creativity is desired and you intentionally give up some control. You can prevent this by writing “use the prompt unchanged as entered.” And if you are not writing in English, “use the prompt unchanged as entered, and only translate it into English.”
When pushing DALL-E to creative limits, nonsensical texts suddenly appear, where DALL-E inserts the prompt into the image, probably to describe it. This has been the strangest behavior so far. You cannot get rid of it with “don’t add any text”, on the contrary, you get more text. You have to change the prompt itself.
DALL-E seems to use some templates for image generation to increase the likelihood of appealing images. Depending on where these are triggered, they are almost impossible or completely impossible to remove.
It is also good to avoid possibility forms like “should” or “could” and instead directly describe what you want to have in the image.
And, write the most important thing first, then the details, and finally technical instructions like image size, etc.
Also, geometries are not yet fully understood. For example, a snake sometimes simply consists of a closed ring. The system is still not perfect…
It is also interesting to know that even GPT does not recognize some of these weaknesses and generates prompts for DALL-E that need to be improved.
Hope this helps.
