How to Install and Use Stable Diffusion with One of the Best Web UIs – InvokeAI

Categorized as AI/ML, Art, Open Source
Save and Share:

InvokeAI is one of the best free and open-source web interfaces for Stable Siffusion, and a strong competitor to Automatic1111, the most feature-rich and famous SD Web GUI to date.

It has some neat features that make it the best choice for certain use cases, such as inpainting and outpainting.

Installation Steps

System Requirements

  • Supported on Linux, Windows, and macOS
  • For Linux: NVIDIA or AMD graphics card
  • For Windows and macOS: NVIDIA card with at least 4 GB of VRAM or an Apple computer with an M1 chip
    • Note: GTX 1080, 1650 and 1660 series video cards are not recommended due to their limitations
  • Requires 12 GB of disk space for machine learning models and dependencies

Download the Installer

Visit Invoke’s GitHub latest release page, scroll down to the “Assets” section and download the zip file with the installer.

Run the Installer

  1. Unzip the file.
  2. Run the appropriate script based on your operating system:
    • If you’re using Windows, double-click on the install.bat script.
    • If you’re using macOS, open a Terminal window, drag the file install.sh from Finder into the Terminal, and press return.
    • If you’re using Linux, run install.sh.
  3. Confirm the installation location for InvokeAI and its image generation model files. Choose a location with at least 15 GB of free memory. If you plan on installing many models, you may need more space.
  4. Wait for the installer to complete. After installation, a script will launch to configure InvokeAI and select a set of starting image generation models.
  5. Locate the installation folder for InvokeAI. It’s not the same as the unpacked zip file directory. The default installation path is ~/invokeai on Linux/Mac systems, and C:\Users\YourName\invokeai on Windows. Inside this directory, you’ll find launcher scripts named invoke.sh and invoke.bat.
  6. Run the appropriate script based on your operating system:
    • If you’re using Windows, double-click on the invoke.bat file.
    • If you’re using macOS, open a Terminal window, drag the invoke.sh script from the folder to the Terminal, and press return.
    • If you’re using Linux, run invoke.sh.
  7. Type 2 to open the browser-based UI, then press enter/return. Wait a minute or two for Stable Diffusion to start up, then launch your browser and go to http://localhost:9090.

Now you are all set to use InvokeAI’s WebUI for stable diffusion. Enjoy experimenting with this powerful tool and see how it compares to Automatic1111!

Text to Image

Invoke AI’s text to image feature allows you to create images by inputting a text prompt. You can customize the number of images, steps (iterations), and scale. The default resolution can be adjusted, and you can choose the seed for image generation. The interface provides helpful tooltips for each setting.

Face Restoration and Upscaling

Invoke AI offers face restoration with two options: GFP-GAN and Code Form. Upscaling is also available, using ESRGAN as the default upscaler.

Generating and Managing Images

As your image is generated, you can view a live render preview. Once completed, you can send the image to the image-to-image tool or the unified canvas, copy the link, or download it. You can also view the image in greater detail using the viewer.

Image to Image

The image-to-image tool allows you to modify existing images. Simply drag an image into the tool and adjust the denoising strength to make changes. The same settings from the text-to-image tab are available here.

Unified Canvas

Invoke AI’s standout feature is the unified canvas. This tool lets you outpaint images, essentially expanding them beyond their initial boundaries. The interface is smooth and user-friendly, with easy navigation and zooming.

To use the unified canvas, drag an existing image onto the canvas and select the area you want to outpaint. You can then generate new images based on the selected area. Accept or discard the results before moving on to the next area.

Seam Correction, Infill, and Scaling

Seam correction options are available to help blend the outpainted areas with the original image. You can also adjust the infill and scaling settings to improve the resolution of the inpainted parts.

Undo and Redo

The undo and redo features are valuable additions to the unified canvas, allowing for easy adjustments and corrections without the need for external tools.

Future Features

Invoke AI plans to introduce a node-based system for image generation and additional post-processing features. A training section for textual inversion and the dream booth is also under development.

Conclusion

Invoke AI’s main strengths lie in its user experience and stability. While it may not have as many features as Automatic1111, its unified canvas and intuitive interface make it an appealing choice. If Invoke AI continues to improve and expand its capabilities, it has the potential to become a front runner in AI image generation tools.

Leave a comment

Your email address will not be published. Required fields are marked *