this is a small. concise workflow for upscaling/refining images and video with the newest Flux model GGUF. Flux RSPO. LOW VRAM users are more than likely able to run this upscaler with 8gb vram. note that the meta batch only renders videos frame by frame so generation times will be hours for video.

images will upscale on my 3070 in about 10 minutes.

Note: sampler steps are set to 20 but this will COMPLETELY CHANGE the images details in cluding face likeness and even entire outfits. To make MINOR DETAILS, use 2-8 steps maximum. You will still see likeness change but at 2 or 3 steps of refinement you will keep a majority of the features intact. Prompting for it helps. The images in the examples are 20 step change comparisons. So you can see hoe much the model takes over based on high step count.

You can also remove the model and use another of your own in place of srpo if you dont like the detail changes it makes The model focuses on realism.