Flux New Model: Kontext
Flux Kontext model provide you to change detail of image by prompt. This is simple workflow
Description
FAQ
Comments (7)
Its working, thanks for that. But its hard to keep the same person.
btw: Why there are 2 input fields?
Thank you for comment so much. About 2 input fields i separate them because it makes easier to prompt by section like, input 1 prompt about character, input 2 prompt about background, However In the end, both will be merged to become one prompt by Join String Node.
It mean you can leave any one slot empty, it doesn't matter.
And if you want to keep person or face try to see this my workflow
https://civitai.com/models/1726265/flux-workflow-multi-character-with-kontext
@tinnapop_ton Ah, good to know. So i will use one field for lora triggers and the other one for the main prompt. really cool. Took a look at your other other workflow. But how this can help me to keep the face and body of my input picture?
Sorry, iam new to all this AI stuff and playing around with this since 3 days. Iam trying to generate pictures of my model to train a lora for it. But its hard to keep different poses. Face and hairs always changing. But your workflow helps me a lot to switch clothing. That works great. btw: you know if i can also use other base models? At the moment iam using the def.context.fp8 you suggested. Tried out 2 other flux models but then its not working anymore. I think this model is limited. Cant get him to make my girl in underwear... :)
@tinnapop_ton Another short question, sorry. For picture input there is the "Load Image" node in your workflow. But he is ignoring the mask feature, because the mask output is not connected anywhere. How i can do that so iam able to use the mask feature? Thank you for your time.
@Smithinator You have 2 way for set underwear to character.
1. If flux-kontext not work about underwear you can use flux redux for swap underwear from other references image, see this my workflow -> https://civitai.com/models/1715642/flux-workflow-swap-anything-with-redux
It can help you swap anything what you want ...XD...
2. Train new Lora underwear style by base on model flux-kontext, this choice i'm not sure because I have never tried it either.
@Smithinator For the character's facial shape, prompt should avoid anything that would suggest a specific character modification, especially in the facial area. It may not be 100% like the original image, but it will at least be 90%++ close.
@Smithinator For mark of Load image. I didn't use it because i don't know how to use mark with flux-kontext. I'm sorry, in the future if I find out how to use. I will update my workflow agian.