AUTOMATIC1111, also known as A1111, is a leading Graphical User Interface (GUI) designed to cater to the needs of advanced users. As the de facto GUI for Stable Diffusion, it comes with a plethora of features, making it an indispensable tool for users. However, due to its extensive feature list, it could be overwhelming for first-timers. This guide aims to simplify the process, providing an easy-to-follow manual to navigate through AUTOMATIC1111’s GUI.
Table of Contents:
- Downloading and installing AUTOMATIC1111
- Text-to-image tab
- Parameters for Image Generation
- Seed and Extra Seed Options
- Image Restoration and Tiling
- Image-to-Image Tab
- Sketching and Inpainting
- Upscaling Images
- Applying Styles
- Checkpoint Models
- Training Settings
- Quick Settings
Downloading and Installing AUTOMATIC1111
AUTOMATIC1111 can be used on various platforms including Windows, Mac, or Google Colab. Also, consider checking the Quick Start Guide to better understand which version of Stable Diffusion best suits your needs.
The Text-to-Image Tab
Upon launching the GUI, the first tab you’ll encounter is the txt2img tab. As its name implies, this tab is primarily used for converting text prompts into images – a fundamental functionality of AUTOMATIC1111.
Basic Usage
For those using AUTOMATIC1111 for the first time, here are the options you might want to customize:
- Stable Diffusion Checkpoint: Select the model you prefer. For beginners, the v1.5 base model is recommended.
- Prompt: Describe the image you want to generate. For instance, “A surrealist painting of a cat by Salvador Dali”.
- Width and Height: Define the size of the output image. For v1 models, at least one side should be set to 512 pixels.
- Batch size: Determine the number of images to be generated each time. Multiple images are advisable for testing a prompt as each result will slightly vary.
After tweaking the settings, press the Generate button. In a few moments, your images will be ready!
Parameters for Image Generation
The txt2img tab hosts a variety of parameters to fine-tune your image generation. Here are some key parameters:
- Stable Diffusion checkpoint: This dropdown menu allows you to select models. You need to place model files in the designated folder.
- Prompt text box: Here, you can specify what you want to see in the images. Be as detailed and specific as possible.
- Negative Prompt text box: Specify what you don’t want in your images. This feature is particularly useful when using v2 models.
- Sampling method: This dictates the algorithm for the denoising process.
- Sampling steps: This determines the number of sampling steps for the denoising process. A higher count generally yields better results but takes longer.
- Batch count and Batch size: These control the number of images generated in each run and the total number of runs, respectively.
- CFG scale: This parameter controls how much the model should adhere to your prompt.
Seed and Extra Seed Options
The Seed value controls the content of the image. Each generated image has its unique seed value.
AUTOMATIC1111 uses a random seed value if it is set to -1.
The Extra Seed menu offers additional options for further customization, including Variation seed and Variation strength.
Image Restoration and Tiling
AUTOMATIC1111 also offers options for Face Restoration and Tiling. The former applies an additional model to restore defects on faces while the latter allows you to create repeating patterns like wallpaper.
Image-to-Image Tab
The img2img tab is where you can perform image-to-image functions. It’s particularly useful for creating new images that follow the composition of a base image.
Sketching and Inpainting
AUTOMATIC1111 offers a Sketch function where you can create your initial picture. This feature is especially useful when you want to modify an existing image. The Inpainting function, on the other hand, is used for regenerating certain parts of an image.
Upscaling Images
The Extra page in AUTOMATIC1111 allows you to upscale an image. It uses AI upscalers that are usually unavailable on your PC, providing a free solution for high-quality image upscaling.
Applying Styles
AUTOMATIC1111 offers various methods to apply styles to the AI-generated images. Some of these include using prompts, checkpoint models, Lora, LyCORIS, embedding, hypernetworks, and the checkpoint merger.
Checkpoint Models
Thousands of custom checkpoint models, fine-tuned to generate various styles, are freely available. These can be found on platforms like Civitai or Huggingface.
Training Settings
The Train page in AUTOMATIC1111 is for training models. Although this guide won’t go into detail, it’s good to know that it supports textual inversion (embedding) and hypernetwork.
Quick Settings
AUTOMATIC1111’s Settings page offers an extensive list of customizations. Here, you can enable custom shortcuts, select default face restoration methods, download and select a VAE, and much more.
In closing, while AUTOMATIC1111 might seem overwhelming initially, with a little understanding and practice, you’ll soon be navigating it with ease. Happy exploring!