Stable Diffusion XL 1.0 Update: Automatic1111 Function

Stable Diffusion XL 1.0

We were really confused after getting and using paid versions of Midjourney, Lexica Art, Devian Art, or any other image generation models and services. 

 But not a problem because,

StabliliyAI has released a new model called Stable Diffusion XL (SD XL) 1.0. This model can generate fantastic images with the least computational memory. 

And the fun part is this is absolutely free with no restriction for getting or facing those frustrating terms and conditions and all.

Not only that It provides you much more control and options which helps you to generate immense beautiful images in just a few seconds.

The new release of Stable Diffusion 1.5 and Stable Diffusion 1.0 is they can generate more detailed images with less computational resources with extremely high pixels.

 As we know Stable Diffusion 1.5 can generate images with 512 by 512 pixels of images but Stable Diffusion 1.0 can generate images in 1024 by 1024 pixels of images. Not only that, its also easier than the previous version to fine-tune the later one.

Options to run SD XL 1.0:-

1. You can now run Stable Diffusion WEB UI using Gradio Library on your PC which needs a good power GPU with a minimum of 4 GB of VRAM. 

They provided some awesome features like:

  •  Text to image
  •  Image2image mode by reverse Euler method of cross-attention control
  • Outpainting
  • Inpainting by Runway ML
  • Color Sketch
  • Prompt Matrix
  • Upscaling with any Aspect Ratio Options
  • Text Prompt  Attention etc.

2. But if you don’t have those powerful resources then also you can run Stable Diffusion XL(SDXL) on Google Colab.

Google Colab has some limitations with free versions where you get almost  12 GB of VRAM and 80 GB of storage for runtime and an extra 15 GB shared with Google Drive which you can connect to Google Colab and store your files.

But, what we think is that Stablity.ai has done great work in minimizing the computational cost when training or fine-tuning the model with humungous data sets needs lots of RAMS, and GPUS.

3. If you don’t want to make your hands dirty in coding then you can also use Stable Diffusion XL 1.0 On the Clipdrop website powered by Stability.ai. 

Clipdrop provides various types of functionalities for working on images like Cleanup, Image Upscaler, Relighting, Replacing or removing background, Sky replacer, Swapping Face, Uncroping, etc.

 Apart from that you can also use their API which provides options like inpainting, background removal, upscaling, removing text, text-to-image, portrait depth estimation, sketch-to-image etc. Their pricing starts with 500 credits which go to 1000,000 credits.

4. Use Stable Diffusion XL 1.0 on the Lexica Art website to generate images for free with some limitations. 

They provide some free 25 credit limits for a month to generate images but if you want to gain and learn how the prompts work in detail on Lexica art then you can take their premium options as well where you can generate multiple stunning images without any hassle.

5. You can also use an API provided by Clarifai.com to run Stable Diffusion XL 1.0 on your local or remote machines.

 Using this you don’t need to give load to your machine and with just simple prompts you can generate beautiful images instantly.  

6. For detailed and image generation, Stability.ai has recommended using the Comify UI which gives you much more functionality so that you can tweak it to get whatever results you want. 

Comfy UI provides you an easy solution if your computer is older where real Stable diffusion models suffer a lot, easy to install, and whatnot. 

If you are familiar with automatic 1111 then this platform feels like you are an old player in this arena. The custom nodes are awesome which are contributed by the stable diffusion community itself. 

Not only this, 

Comfy UI packs with multiple great extensions like Comfy UI Impact Pack, Effeciaency Node for Comfy UI, WAS Nodes Suite, Ultimate SD Upsclaer, etc. which help you to get an extra edge in the field of image generation over others. 

How to download SDXL 1.0 base model for Automatic1111?

0. First, you need to get installed the Stable Diffusion WebUI(Automatic1111) and upgrade it.

sdxl1.0 base model

1. Then, download the SDXL 1.0 base model by moving to the official hugging face repository from StabilityAI.

Download sdxl1.0 base model

2. Click to “Files and Versions” section, and download the “sd_xl_base_1.0.safetensors” file.

saving sdxl1.0 base model

3. Move to the “stable-diffusion-webuimodelsStable-diffusion” directory and save the downloaded “sd_xl_base_1.0.safetensors” file into the same.

vae download

downloading vae

4. Now, you need to get the Vibrational Auto Encoder (VAE) file by clicking on the “vae” and select “diffusion_pytorch_model.safetensors” file to download it. Then rename it to something that gets easily identified while working with WebUIs. We have renamed it to “sdxl_vae.safetensors“.

If you are a windows user, then you need to enable the option “File name extensions” under Menu section otherwise you can’t rename the files with its extensions.

Saving Vae

5. Save the downloaded file for VAE to the “stable-diffusion-webuimodelsVAE” folder.

5. Again, you need a SDXL Refiner 1.0. So, move to the official hugging face page provided by StabilityAI.

downloading sdxl refiner 1.0

6. Click to “Files and Versions” section, and download the “sd_xl_refiner_1.0.safetensors” file. The SDXL refiner is used to add the detailing for generating higher resolution images.

saving sdxl1.0 base model

7. Again, move to the “stable-diffusion-webuimodelsStable-diffusion” folder and save the “sd_xl_refiner_1.0.safetensors” model into it.

8. Now, restart your Automatic1111 to take effect.

You can generate tones of AI prompts using our Stable Diffusion Prompt Generator for Stable Diffusion XL(SDXL) models.

Conclusion:

Stable Diffusion XL a gigantic model trained on huge dataset is really a game changer model for AI community. Due to its large dataset one can train LoRA models, fine tune their models in a better manner. This model is capable to run in Automatic1111, ComfyUI, Fooocus, SwarmUI and even in various cloud as well.