bye midjourney! SDXL 1.0 - How to install Stable Diffusion XL 1.0 (Automatic1111 & ComfyUI Tutorial)
TLDRThe video tutorial introduces the installation and use of Stable Diffusion XL 1.0, a free AI tool for generating diverse and detailed images. It covers three methods: local installation with Automatic1111 and ComfyUI, and online use via clipdrop.co. The tutorial demonstrates generating images with different styles and resolutions, highlighting the benefits of using the refiner model for enhanced detail. It also mentions upcoming content on advanced features.
Takeaways
- 🎉 Stable Diffusion XL 1.0 (SDXL 1.0) is released and can generate various styles of images for free.
- 🖼️ The software can create realistic and anime-style images, as well as other interesting styles.
- 🛠️ There are three methods to use SDXL 1.0: locally with automatic1111, locally with ComfyUI, and online for free.
- 📥 To run locally, download the base model and optionally the refiner model from the Hugging Face repository.
- 🔗 Use the pre-built file from the official GitHub repository for automatic1111, no need to install Python or git.
- 🔄 Extract and copy the SDXL base model into the specific folders within the automatic1111 webui directory.
- 💻 Run scripts to update the code and install dependencies, then start the server to use the UI.
- 🖼️ For better image generation, use the recommended 1024x1024 resolution or other natively supported resolutions.
- 🌐 If automatic1111 is not suitable, ComfyUI offers support for the refiner model and different UI experience.
- 🔄 In ComfyUI, paste the models into the checkpoints folder and run the appropriate batch file to launch the UI.
- 🔍 The refiner model can be added to ComfyUI for more detailed image generation, enhancing faces, backgrounds, and textures.
- 🌐 For those unable to run SDXL locally, clipdrop.co offers a free online alternative with a generation limit.
Q & A
What is Stable Diffusion XL 1.0 (SDXL 1.0) and what can it do?
-Stable Diffusion XL 1.0 is a model that can generate various styles, types, subjects, and backgrounds of images for free. It is capable of producing realistic and anime-style images, among other styles.
How many methods are presented in the script for using SDXL 1.0?
-Three methods are presented: running it locally using automatic1111, running it locally with ComfyUI, and using it online for free.
What is the first step to run SDXL 1.0 locally using automatic1111?
-The first step is to download the SDXL 1.0 base model and optionally the refiner model from the huggingface repository by Stability AI.
Why is the pre-built file of automatic1111 used instead of installing Python or Git?
-The pre-built file is used for convenience and to simplify the process, eliminating the need for users to install Python or Git.
How do you update the code to the latest version that supports SDXL in automatic1111?
-By running the 'update.bat' script, which updates the code to the latest version supporting SDXL.
What is the default server address for the automatic1111 UI after installation?
-The default server address is localhost:7860.
What is the recommended image size for generating better results with SDXL 1.0?
-The recommended image size is 1024x1024, as SDXL was primarily trained on this size.
Why might ComfyUI be a better choice for some users?
-ComfyUI might be a better choice for users who experience issues with automatic1111, need a faster interface, or want to use the refiner model.
How do you add the refiner model in ComfyUI?
-You load the refiner model by clicking on 'load' in the refiner's load checkpoint section and selecting the refiner.json file, ensuring the actual refiner model is loaded in both the refiner and base model checkpoints.
What is the online alternative mentioned for users who cannot run SDXL 1.0 locally?
-The online alternative is clipdrop.co, a product by Stability AI, which allows free image generation with a limit of 400 generations.
What additional topics will be covered in future videos according to the script?
-Future videos will cover topics such as LoRA training, Dreambooth, ControlNet, and running SDXL on low VRAM devices.
Outlines
🖼️ SDXL 1.0 Launch and Local Installation Guide
The script introduces the release of SDXL 1.0, a tool for generating diverse images in various styles and subjects for free. It outlines three methods for using SDXL: locally with automatic1111 UI, with comfy UI, and online for free if local options are not viable. The detailed guide for the first method includes downloading the base and optional refiner models from Stability AI's huggingface repository, using a pre-built file from the official GitHub repository, and setting up the environment by extracting files, updating scripts, and installing dependencies. The script also demonstrates image generation with positive and negative prompts, emphasizing the optimal use of 1024x1024 resolution for better results.
🔧 Exploring ComfyUI and Online Alternatives for SDXL 1.0
This paragraph focuses on using ComfyUI as an alternative to automatic1111 for running SDXL 1.0 locally, especially when wanting to utilize the refiner model. It provides a step-by-step guide for downloading and setting up ComfyUI, including pasting the base and refiner models into the correct folders and launching the interface. The script illustrates the process of image generation with different prompts and settings, highlighting the refiner's ability to add details to the generated images. It also mentions an online alternative, clipdrop.co, for those unable to run SDXL locally, and hints at future tutorials on advanced features and optimizations for low VRAM devices.
Mindmap
Keywords
💡Stable Diffusion XL 1.0
💡Automatic1111
💡ComfyUI
💡Refiner Model
💡Huggingface Repository
💡Localhost
💡Positive Prompt
💡Negative Prompt
💡Resolution
💡Sampler
💡Clipdrop.co
Highlights
SDXL 1.0 is released and can generate a wide variety of images in different styles, types, subjects, and backgrounds for free.
The tutorial provides three methods to use SDXL 1.0: local installation with Automatic1111, ComfyUI, and online use.
To use SDXL 1.0 locally, you need to download the base model and optionally the refiner model from the Hugging Face repository.
Automatic1111 is the most popular UI for Stable Diffusion and requires downloading a pre-built file from its GitHub repository.
No need to install Python or Git for Automatic1111 as it uses a pre-built file for installation on Windows.
The SDXL base model should be placed in the 'Stable-diffusion' folder within the Automatic1111 webui/models directory.
Running the 'update.bat' script updates the code to the latest version supporting SDXL.
After installation, the server starts on localhost:7860, and the web UI can be accessed through the browser.
SDXL generates better images at its primary training size of 1024x1024, which may increase generation time.
ComfyUI offers the best support for SDXL since launch and natively supports the SDXL refiner.
ComfyUI can be downloaded from its official GitHub repository and requires extraction of the 7zip file.
SDXL 1.0 base and refiner models should be placed in the 'checkpoints' folder of ComfyUI's 'models' directory.
ComfyUI can be launched with a batch file depending on whether you have an NVIDIA GPU or not.
The refiner model in ComfyUI can add more details to the generated images, enhancing face, background, and minor details.
If local installation is not possible, SDXL can be used online for free at clipdrop.co with a limit of 400 free generations.
The tutorial promises future videos on running SDXL on Colab and other advanced features like LoRA training and Dreambooth.
The presenter is currently unwell but plans to return with more content on SDXL once recovered.