Possible to use SDXL or FLUX models instead of SD2.1?
I notice a custom pipeline is being used using the diffusers library. Can SDXL pipeline be used or better yet, FLUX models be used instead of the Stable Diffusion 2.1 (SD2.1)?
SD2.1 is considered legacy model, while CCSRv2 offers great results the SD2.1 is a limiting factor.
Did you try with SD 3.5 ?
Thank you for your feedback! Currently, CCSR does not support other SD models. However, SD 2.1 can be directly replaced with other SD models within the Diffusers library. Since part of CCSR's generative capability is supported by the SD decoder with the optimization of GAN loss, replacing the base model may not provide significant advantages in a single step. We welcome everyone to experiment and share their feedback!