OPENGIT/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2025-01-31 02:32:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

add sdxl model dl link update + refiner info with ext. included

ClashSAN 2023-07-29 00:27:27 -04:00
parent e53c309b20
commit a60925601b
1 changed files with 16 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
Features.md

@ -3,13 +3,22 @@ This is a feature showcase page for [Stable Diffusion web UI](https://github.com
All examples are non-cherrypicked unless specified otherwise.
# SD-XL
## SD-XL BASE
## Downloading:
1. [sd_xl_base_1.0_0.9vae](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0_0.9vae.safetensors)
2. [sd_xl_refiner_1.0_0.9vae](https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/blob/main/sd_xl_refiner_1.0_0.9vae.safetensors)
These models are recommended for generating, merging and training.
>They have a built-in trained vae by https://github.com/madebyollin which fixes NaN infinity calculations running in fp16. ([Vae](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl_vae.safetensors) ) for reference
## SD-XL BASE
[[PR]](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/11757) | [[Stability-AI Github]](https://github.com/Stability-AI/generative-models)
### Info
This is a model designed for generating quality 1024×1024-sized images. It is not meant to generate good pictures at 512×512.
This is a model designed for generating quality `1024×1024`-sized images. It is **not** meant to generate good pictures at `512×512`.
It's tested to produce same (or very close) images as Stability-AI's repo (need to set Random number generator source = CPU in settings)
@ -30,18 +39,16 @@ It's tested to produce same (or very close) images as Stability-AI's repo (need
- works at minimum 4gb gpu (30XX)
```
>NOTE: Initial loading of these models require those with 24gb cpu memory (RAM) and under to setup a large pagefile. This may change in the future.
>NOTE: Initial loading of these models require those with 24gb cpu memory (RAM) and under to setup a large pagefile. This may change in the future .
### Downloading:
1. [sd_xl_base_1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0.safetensors)
## SD-XL REFINER
>The base model's VAE has problems running in fp16. [madebyollin](https://github.com/madebyollin) has kindly trained a vae to mostly remedy this:
This secondary model is **designed** to process the `1024×1024` SD-XL image **near completion***, to further enhance and refine details in your final output picture. You could use it to refine finished pictures in the img2img tab as well.
2. [sdxl-vae-fp16-fix](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl_vae.safetensors)
(Please use both of these to avoid slow speeds, and GPU out-of-vram issues.)
To try this kind of generation, you can use this extension - https://github.com/wcde/sd-webui-refiner
You will want 6gb minimum as `--lowvram` currently isn't supported.
# SD2 Variation Models
[PR](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/8958), ([more info.](https://github.com/Stability-AI/stablediffusion/blob/main/doc/UNCLIP.MD))