Stable Diffusion with Automatic 1111 on Jarvislabs

Published in

Jarvislabs.ai

4 min readJan 2, 2023

Generating images with Stable diffusion(SD) has gotten super easy over the last few months. Automatic 1111 is a popular open-source UI tool built to help enthusiasts and artists play around with SD and apply many advanced features.

In this blog post, we will delve into, how you can use Automatic1111 a powerful platform on Jarvislabs.ai in a few clicks and transform your creative process. Create stunning visuals with ease, with 10+ preloaded models.

What is AUTOMATIC1111 — Stablediffusion UI?

Automatic 1111 provides an open-source UI dashboard built on top of Gradio. With more than 350 active contributors, it brings in some of the latest advancements in Generative AI space to its users.

Key features of Automatic 1111:

txt2img and img2img modes
Inpainting and Outpainting
Textual Inversion
CLIP interrogator, a button that tries to guess a prompt from an image

There is a ton of features available and they are growing. Check here for features with examples.

Setting up AUTOMATIC1111 on JarvisLabs.ai

Getting started with AUTOMATIC1111 is a breeze. Get started in less than 60+ seconds ⏰. It comes with 10+ preloaded models.

Launch an instance, choose

Choose Automatic 1111 as a framework
GPU Typ — (Recommended A5000, A6000)
Storage — (You can increase or decrease it anytime)
Number of GPUs — (1 GPU should be ideal for most of the cases)

Easily access the AUTOMATIC1111 application by right-clicking on the instance and selecting the API endpoint for a smooth creative experience on JarvisLabs.ai.

Note: The UI may need a few seconds to load, please refresh the UI if you don’t get the UI.

Keep an 👀 on the docs for updated information.

Exploring with 10+ Preloaded Models

The first step in using Automatic 1111 is to choose the right model and place it in the right location. We have done the hard work of picking some of the most popular models and have placed them in the Automatic 1111 models directory. So you can focus on Image generation.

You can choose from one of the popular models and start generating images.

Download model weights and config

It is super simple to add any new models or custom pretrained models. You can follow these simple steps.

Once your instance is launched, clicking the play button will start a Jupyter Lab.

From the Jupyter lab, let's open a terminal to download the model weights and config.

Download your favorite model from places like Hugging face or civitai and place it under

/home/stable-diffusion-webui/models/Stable-diffusion/

To try SD 2.1, we need to download the weights and config.

#Download weights
wget https://huggingface.co/stabilityai/stable-diffusion-2-1/resolve/main/v2-1_768-ema-pruned.ckpt

#Download config
wget https://raw.githubusercontent.com/Stability-AI/stablediffusion/main/configs/stable-diffusion/v2-inference-v.yaml

Let's rename the config file and place it in the model's directory.

mv v2-inference-v.yaml v2-1_768-ema-pruned.yaml
mv v2-1_768-ema-pruned* stable-diffusion-webui/models/Stable-diffusion/

You do not need a config file for all the models.

What's next

In the coming weeks, I will create blogs that show

How you can train models that create images of you, your pet, or any object.
How to use a technique called Textual Inversion to generate images in different styles.
How to upscale generated images and more.
Try another popular UI for SD called InvokeAI

I hope you find the blog useful, would love to know your thoughts on how we can improve it. In case you have any suggestions/challenges you can comment or ping us on the chat available on Jarvislabs.ai