Background images without objects
I am using this model to create background images, that can be used for various purposes (i.e. putting an object on the background image later on).
However, the model generates product images in the final picture as well (I only want a background picture, without any objects).
For example, the prompt that I use is "A good background image that can be used for advertising mugs. Only provide the background image, do not include any other objects"
But the image actually includes mugs (which I do not want):
Or in another example, the prompt is "A good background image that can be used for advertising TVs. Only provide the background image, do not include any other objects"
but again, the image contains a TV:
Can you please advise why the model puts objects in the image even though in the prompt we ask not to?
Thanks
Because that isn't how you prompt an image generation model yet. It can't think or do any complex calculation to determine anything ambiguous in your prompt. In general it can't follow instructions either.
Simply describe what you want as if it were the caption of the image you are trying to create. "A photo of an modern living room with grey wood panel floor, warm lighting". The more words (tokens) you use, the less it will follow each word, so use as few as you can get away with. Order the most important things in the picture from left to right in your prompt. Put things in the negative prompt that you don't want IF they show up in the image generated. "Furniture, painting, lamp", etc. You can find many tutorials on how to prompt SDXL on youtube etc. if you need more help.