Flux1 Dev GGUF 8GB VRAM ControlNet Upscaler/Refiner. - v2.0 Showcase

Flux ControlNet image upscaler V2 (with refine support) with tweaked params for minor image changes/corrupt, by default for 8GB VRAM, but also can be modified for non-GGUF flux models.

Workflow: Input image / Output image.

Models:

flux dev gguf: https://huggingface.co/city96/FLUX.1-dev-gguf/resolve/main/flux1-dev-Q4_K_S.gguf

flux controlnet: https://civitai.com/api/download/models/901095?type=Model&format=SafeTensor

flux vae: https://huggingface.co/f5aiteam/VAE/resolve/main/ae.safetensors (rename to flux1_vae.safetensors).

clip 1: https://huggingface.co/f5aiteam/CLIP/resolve/main/clip_l.safetensors

clip 2: https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/resolve/main/t5-v1_1-xxl-encoder-Q8_0.gguf

soft upscaler: https://huggingface.co/Acly/Omni-SR/blob/main/OmniSR_X4_DIV2K.safetensors

Additional models for V2:

new upscaler: https://huggingface.co/uwg/upscaler/resolve/main/ESRGAN/4x_NMKD-Siax_200k.pth

pre/post process upscaler 1: https://huggingface.co/utnah/esrgan/resolve/main/1x_NMKD-Jaywreck3-Soft-Lite_320k.pth

pre/post process upscaler 2: https://huggingface.co/utnah/esrgan/resolve/main/1x_NMKD-Jaywreck3-Lite_320k.pth

lora: https://civitai.com/api/download/models/964759?type=Model&format=SafeTensor

Notes:

By default configured for upscaling/finetuning normal quality images to 1080p resolution.

Deis sampler was chosen because in my opinion it gives less polished results, with more realistic imperfections and skin (compared to euler or dpmpp_2k), but you need test yourself.

Potentially it may be enhanced by faces/hands enhancing workflows (with florance2) and become even better.