Here is how to install Stable Diffusion and a few tips. This install has one extra step from the older v1.5 install but it's will only take you 10 seconds extra.
Perhaps the highlight of the show is Stable Diffusion's ability to do Depth to Image mapping which is similar to when you add textures to a polygon, but this will be done with Image prompting onto an exiting model to serve as a reference which is wild and is something that now separates Stable Diffusion from the rest of the Ai Art generators. Image what game development will become with this tech, we might be able to generate things on the fly while in game or create characters simply by prompting onto a static mesh or polygon. Amazing!!!
Version 2.0 Information
Here is the official Stability AI info on the New 2.0 Release.
Here is Automatic 1111 Features and Showcase link.
Necessary Files
Here is the Web UI Automatic 1111 Repository Files you will need.
Here is the Required Models CKPT file (The Big Boy 5GB)
Here is a download for the Config File down below. Alternatively, you can just grab the from this link and paste it in a notepad and name that file 768-v-ema.yaml.txt and put that file in your Stable Diffusion\stable-diffusion-webui\models\Stable-diffusion . **Hint** this is the same folder you put your Checkpoint files in.
Here is the newest step you will have to right click here and then save link as and save it to your 768-v-ema.yaml in your Stable Diffusion\stable-diffusion-webui\models\Stable-diffusion. **Hint, this is the same folder that you put your CKPT files in.
You can also find the config file link from the features site at the top.
1st timers! You will need to download Python 3.10.6 (DO NOT Download the newest version of python)
AMD User? Click here.
Download and install Python 3.10.6 (Direct Link to 64bit installer for windows) Here is the link to the Python Website. IMPORANT! Click the option "Add Python to Path" when asked. If you didn't see this and your webui-user.bat file won't run. Reinstall python > Modify > Add python to path.
Download and install GIT, just say next to everything to install this.
Download Automatic 1111 zip files, from the GitHub repository site by clicking on the green code button drop down arrow.
4. Once you have downloaded that file just extract and unzip those files anywhere in your PC that you can easily access.
If you installed this before, just unzip it to you current Stable Diffusion folder.
5. Download the 768-v-ema.ckpt checkpoint file from the Hugging face download link and
then put that file into your Stable Diffusion\stable-diffusion-webui\models\Stable-diffusion
6. Grab the config file. To do that, you will have to right click here and then save link-as and save it as 768-v-ema.yaml to your Stable Diffusion\stable-diffusion-webui\models\Stable-diffusion.
**Hint, this is the same folder that you put your CKPT files in.
Once you unzipped the new Automatic 1111 Webui respository have your Checkpoint 768-v-ema.ckpt (5GB file) and your 768-v-ema-.yaml config file in your models / Stable Diffusion folder your ready to run the Webui-user.bat file. So double click on it and it will install dependences for 10 minutes. **When you see the words "loading weights" on your cmd console your PC might freeze up, try to give it sometime to finish loading everything and it is almost done.
Once the Webui-user.bat file is done installing the dependencies you will see the message "Running from Local Url: 127.0.0.1:7860 you can copy and paste that Url IP address (local IP address and port 7860) and put that in any browser. Then you will have Stable Diffusion 2.0 up and running.
One important last step is you will have to select your checkpoint file by clicking on thedrop-down box at the top left of your Stable Diffusion, use the drop-down box and select
the 768-v-ema.ckpt file.
This will actually prompt your command console to start downloading dependencies so it will take some time as there is 3.9GB of info it is downloading and installing.
Once that is done you can start prompt again like normal.
I will leave you with this... Stable diffusion seems to have much heavier weights when it comes bad images right now and automatic 1111 takes out some of those issues but this is a different diffusion model, and it will not react the same that your probably used to. It is very important to add negative prompt weights. Copy and paste this info in your negative prompt to get some better results:
•Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ((((mutated hands and fingers)))), (((out of frame)))
Once you put your negative prompts in, make sure to save
Just think about it logically or add prompts as needed. If your doing realistic artwork you will have to add negative prompts for cartoonish, watercolor, abstract and things that might make your picture look less real. The effect of negative prompts is much greater in version 2.0.
Here is a youtube video from a "Different youtuber" showing examples on how negative prompts can improve your results and what you images look like before and after.
Need Some Examples of Ai Art Prompts? Try Lexica Ai, you can browse through some of the pictures and click on them to see what prompts were used to generate that picture to understand what art style and words were used to achieve that result. CAUTION its NSFW
Comentarios