QWEN EDIT 2511 seems to be a downgrade when doing small edits with two images.
Been doing clothes swaps for local shop so I have 2 target models (male and female) and I then use the clothing images from their supplier. I could extract the clothes first but with 2509 it's been working fine keeping them on the source person and prompting to extract the clothes and place them on image 1.
BUT, with 2511, after hours of playing, it will not only transfer the clothes (very well) but also the skin tone of the source model! This means that the outputs end up with darker tanned arms or midrif than the persons original skin!
Never had this isssue with 2509. I've tried adding things like "do not change skin tone" etc but it insists on bring it over with the clothes.
As a test I did an interim edit of converting the original clothing model/person to gray manniquin and guess what, the person ends up with gray skin haha! Again, absolutely fine with 2509.
https://redd.it/1pxzbh2
@rStableDiffusion
Been doing clothes swaps for local shop so I have 2 target models (male and female) and I then use the clothing images from their supplier. I could extract the clothes first but with 2509 it's been working fine keeping them on the source person and prompting to extract the clothes and place them on image 1.
BUT, with 2511, after hours of playing, it will not only transfer the clothes (very well) but also the skin tone of the source model! This means that the outputs end up with darker tanned arms or midrif than the persons original skin!
Never had this isssue with 2509. I've tried adding things like "do not change skin tone" etc but it insists on bring it over with the clothes.
As a test I did an interim edit of converting the original clothing model/person to gray manniquin and guess what, the person ends up with gray skin haha! Again, absolutely fine with 2509.
https://redd.it/1pxzbh2
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit
Explore this post and more from the StableDiffusion community
People who are using llm to enhance prompt, what is your system prompt?
I mostly interested in a image, will appreciate anyone who willing to share their prompts.
https://redd.it/1pxzyfc
@rStableDiffusion
I mostly interested in a image, will appreciate anyone who willing to share their prompts.
https://redd.it/1pxzyfc
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit
Explore this post and more from the StableDiffusion community
Semantic Image Disassembler (SID) is a VLM-based tool for prompt extraction, semantic style transfer and re-composing (de-summarization).
https://redd.it/1py4bxf
@rStableDiffusion
https://redd.it/1py4bxf
@rStableDiffusion
Reddit
From the StableDiffusion community on Reddit: Semantic Image Disassembler (SID) is a VLM-based tool for prompt extraction, semantic…
Explore this post and more from the StableDiffusion community