I've been thinking, what if, instead of porting concepts like Tryon and Face adapter, we didn't create a new concept that, in addition to this fit, would be widespread.
I arrived at this concept with your repository with ip adapter and inpaint.
I saw that your model also supports image as input in the ip adapter, in addition to the projection of the initial image (I believe it is something for img2image), but what if in addition to this input we also add the mask input input to generate the content in the conditioned mask by inpaint?
Something like loading the flux with the IP adapter features and recreating the item in an input image with inpaint?
I've been thinking, what if, instead of porting concepts like Tryon and Face adapter, we didn't create a new concept that, in addition to this fit, would be widespread. I arrived at this concept with your repository with ip adapter and inpaint. I saw that your model also supports image as input in the ip adapter, in addition to the projection of the initial image (I believe it is something for img2image), but what if in addition to this input we also add the mask input input to generate the content in the conditioned mask by inpaint?
Something like loading the flux with the IP adapter features and recreating the item in an input image with inpaint?
im calling it flux carrier