Stable Cascade provides the advantages of exceptional quality coupled with unrivalled resolution. Stable Cascade Course: - cutt.ly/xwBnvtjV Stable Cascade playlist kzbin.info/aero/PL4nokMsvRs7vXbLiBgsDuSEnVexKWXsMe Beginner's Guide to ComfyUI and Stable Diffusion KZbin - bit.ly/BeginCOMFYUISDXL Udemy - bit.ly/SDGenStart Join 🏆 this channel to get access to exclusive content and perks: bit.ly/XOVjoin
@LouisGedo11 ай бұрын
👋
@Adreitz711 ай бұрын
I need to see if this works on my Mac through Pinokio. I currently enjoy generating with SDXL in Invoke using an upscaling workflow to get up to 4k+ (8.8MP) resolution. It generally works pretty well, but the refiner (which in my opinion is required for quality images) kind of goes crazy at high resolutions even with the initial generation for guidance, leading to a certain patterned look across parts of the image that are supposed to be smooth. Generating directly at high res with SDXL, from my limited experiments, tends to change the image composition in ways I don't like, though it is better than SD 1.5 in that regard. Have you noticed any problems with composition in your high res generations? I've also read some concerns about the latent compression factor with Cascade possibly leading to it producing images that look a lot more samey. Any comments about that?
@b4ngo54010 ай бұрын
it seems like this model has an issue with long necks, most of the images I saw online were with necks longer than average, also most of my tests got long necks too try increasing the latent resolution a little bit and the neck will go really long like a giraffe
@robadams245111 ай бұрын
The detail fur and stuff are very nice but I find the faces poor. Like plastic waxworks with zero character. The Dalle faces are worse though, like incompetent airbrushing. The impressionist things looked hopeful.