AI painting artifact: SD XL 1.0 heavy update! Free open source and commercially available (with online use + local deployment tutorial)
1. Use SDXL 1.0 online for free
1.1. Discord serverSD XL1.0 Discord server: https://discord.com/invite/stablediffusion
Click the link above to join the SD XL1.0 official server, click any channel in "SDXL BETA BOT", enter /Dream in the text box, then enter the text prompt, press Enter to generate it. The Wenshengtu function can be used for free at present, but there will be a limit on the frequency of use. When the limit is limited, you can wait for a while or try another channel.
1.2. Clipdrop
Direct website: https://clipdrop.co/stable-diffusion
Clipdrop is an online AI image processing tool acquired by Stability AI. Now it also has the latest SDXL 1.0 model built-in, and has added 2 new functions of screen aspect ratio and negative prompt words. The generation service is also free to use, but there will be a Clipdrop watermark in the lower right corner of the image.
2. Deploy SDXL 1.0 locally
2.1 Install Python 3.10.6 (newer versions of Python do not support torch), check "Add Python to PATH".
2.2 Install git.
2.3 Download stable-diffusion-webui 【Click to download】
2.4 If your computer's video memory is less than or equal to 8G, please edit the webui-user.bat file and add the following parameters after set COMMANDLINE_ARGS=: –medvram (video memory below 8G) or –lowvram (video memory below 4G).
2.5 Download 2 large models, 1 Lora model and 1 Vae model. The large model is installed to the models\Stable-diffusion folder in the root directory, the Lora model is installed to the models\Lora folder, and the Vae model is installed to the models\VAE folder (there is a resource package at the end of the article).
a. Base large model and lora model: stabilityai/stable-diffusion-xl-base-1.0 at main
b. Refining the large model: stabilityai/stable-diffusion-xl-refiner-1.0 at main
c. Vae model: stabilityai/sdxl-vae at main
Comments
Post a Comment