SD Compendiumhttps://www.sdcompendium.com/doku.php?id=start
Extensive wiki about Stable diffusion.
Extensive wiki about Stable diffusion.
A mini wiki or cheat sheet for Stable Diffusion. The purpose of the mini wiki is to provide a brief explanation and relevant links to tools that could help users generate the exact image they want. The focus is on how to obtain the desired image rather than how Stable Diffusion works. The page aims to address common problems and gaps in knowledge, and welcomes suggestions from users to improve the resource.
A list of custom ESRGAN models.
Enhanced SRGAN (ESRGAN) is an improvement of the Super-Resolution Generative Adversarial Network (SR-GAN) that is capable of generating realistic textures during single image super-resolution. ESRGAN improves three key components of SRGAN, including network architecture, adversarial loss, and perceptual loss, to enhance the visual quality of images. ESRGAN introduces the Residual-in-Residual Dense Block (RRDB) without batch normalization as the basic network building unit, borrows the idea from relativistic GAN to let the discriminator predict relative realness, and improves the perceptual loss by using the features before activation for stronger supervision. ESRGAN achieves better visual quality with more realistic and natural textures than SRGAN, winning the first place in the PIRM2018-SR Challenge.
Stable Diffusion is an open-source diffusion model used for generating images from textual descriptions. The input parameters for generating images include prompt, negative prompt, sampling steps, sampling method, ddim_eta, batch count/n_iter, batch size, CFG Scale, width, height, and seed. Increasing batch size and width requires more VRAM, while increasing CFG Scale results in images that resemble the prompt more closely but degrades image quality after a certain point. The seed is the starting point for the RNG, and varying input parameters only slightly can still produce good results using previously successful seeds.