Soft images

[witotoole]

Using StableSR I get soft images with little detail, compared to Ultimate SD Upscale.

StableSR to the leflt, Ultimate SD Upscale to the right of the comparison

This is the prompt used

4K, Masterpiece, highres, absurdres,natural volumetric lighting and best shadows, smiling,nikon d850, 85mm F1.2mm lens portrait of pretty woman posing for the camera

Negative prompt: ((disfigured)), ((bad art)), ((deformed)),((extra limbs)),((close up)),((b&w)), wierd colors, blurry, (((duplicate))), ((morbid)), ((mutilated)), [out of frame], (((extra fingers, mutated hands))), ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))), out of frame, ugly, extra limbs, (bad anatomy), gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), mutated hands, (fused fingers), (too many fingers), (((long neck))), Photoshop, video game, ugly, tiling, poorly drawn hands, poorly drawn feet, poorly drawn face, out of frame, mutation, mutated, extra limbs, extra legs, extra arms, disfigured, deformed, cross-eye, body out of frame, blurry, bad art, bad anatomy,(((UGLY DEFORMED, NDS))),(((black and white picture))),(((cat ears))),(((wings))),(((cleavage,boobs))),young,loli,child

Steps: 20, Sampler: Euler a, CFG scale: 2, Seed: 2894591607, Size: 1840x3680, Model hash: ac34765554, Model: icbinpICantBelieveIts_v6, Denoising strength: 0.05, Clip skip: 2, ENSD: 31337, Version: v1.2.1, Tiled Diffusion: "{'Method': 'Mixture of Diffusers', 'Tile tile width': 64, 'Tile tile height': 64, 'Tile Overlap': 32, 'Tile batch size': 8, 'Keep input size': True}", Eta: 0.67

I notice that the higher the denoise setting, the softer the image

[pkuliyi2015]

For StableSR you must use the sd2.1 512 official model. It doesn't support anything else.

https://huggingface.co/stabilityai/stable-diffusion-2-1-base/tree/main

Download the ema one.

[witotoole]

So use it here?

[pkuliyi2015]

So use it here?

You are right.

[witotoole]

Images still a bit soft, but will play more with it.