Image Generator: MitchJourn-E

Introducing MitchJourn-E! The text-to-image generator app for Windows. This app aims to enable intuitive, fast, and exciting ways to generate new images, using the power of InvokeAI and Stable Diffusion on Windows desktop.

What is MitchJourn-E? The name is a play on words, referencing the popular image generation services “MidJourney” and “Dall-E”. This app provides an easy-to-use desktop interface for the popular Stable Diffusion fork called “InvokeAI“. MitchJourn-E is an open-source Windows desktop application that allows you to type in a description of an image and the app will create that image using the Stable Diffusion model.

A screenshot of the MitchJourn-E application showing an image generated from the prompt "a beautiful scenic landscape of a Hawaiian sunset"
A screenshot of the MitchJourn-E application

Download for Free

MitchJourn-E is an open-source free download available from GitHub. Click the green “Code” button and choose to clone or download as a zip. Within the downloaded folder, run MitchJourn-e.exe.

Generated Image Examples

Now I'm fucked up and I'm missing you

Features

Text-to-Image: enter text prompts and generate an image

Image-to-Image: reference an existing image to create images with similar composition, style, and size

Create variations: reference a previously generated or local image and create similar but different variations of it

Negative prompting: enter prompts that are banned from generation, usually generating inverse results. Combine with regular and image prompting to fine-tune results

Prompt blending: define two separate weighted prompts to generate an image of intersecting concepts

Prompt Helper: a customizable menu to quickly add commonly used phrases, artist names and styles to speed up workflow

Continuous prompting: Images can be generated one after another, while changing the prompt and the parameters of the prompt to fine-tune the image

Sequential prompting: Images can be generated based on the previous image, allowing prior results shape future results

Upscaling: generate images quickly at low resolution, then upscale them after, for an overall fast way to generate full sized images

High Resolution generation: Generate images at resolutions higher than usually possible, using a 2-phase approach of generating a smaller image, upscaling it, then using the upscaled image as a reference, generate the high-resolution image

Aspect ratio presets: define presets (or use the defaults) to create images in common ratios like 16:9, 2:3, 1:1

Creativity slider: quickly modify multiple parameters with one slider movement that will result in more and less “creative” results, useful for moving between photo-realism and fantasy

Prompt noise: add Perlin noise to the prompt to subtly (or drastically) modify the starting point of the image generation

Latent limiter: limit the amount of data available to the generation, which can generate more photo-realistic/true to source images

Seamless tiling: generate images that can seamlessly tile together, great for clothing patterns and game assets

Generate progress images: save every step the generation makes before creating the end result (for you noise lovers!)

Output sorting (experimental feature): automatically sort generated images into folders based on the prompt used

Prompt folder (experimental feature): reference a folder of images and generate new images referencing each image in that folder sequentially. Good for creating videos

Out-painting (experimental feature): generate additional resolution to existing images, based on the prompt

Metadata Extractor (experimental feature): reference an image and view all the associated metadata. Typically, images generated with Stable Diffusion have the prompt and other parameter info saved within the metadata

Installation

  1. Download and install InvokeAI to the root of the C drive. Follow this separate setup guide (easy) to install Anaconda, Python, and all the other prerequisites automatically: https://invoke-ai.github.io/InvokeAI/installation/INSTALL_INVOKE/#installation
    • When complete, you should have a folder structure like C:\InvokeAI-Main\
  2. Now you need to download a model file. Here is Stable Diffusion 1.5 (v1-5-pruned-emaonly.ckpt) (recommended for best general results). You will need to create a hugging face account before it shows the download link: https://huggingface.co/runwayml/stable-diffusion-v1-5
  3. Rename the .ckpt file to “model.ckpt” and place it here: C:\InvokeAI-main\models\ldm\
  4. Now relaunch MitchJourn-E, and you can start generating images!

Where credit is due

Thanks to Emad Mostaque and his team for creating the Stable Diffusion platform and model which this application uses as it’s backbone. Additionally, thanks to Lincoln Stein and contributors for creating the Stable Diffusion fork which this application leans on for the majority of its image processing features.

This application is a user interface that leverages the hard work of many creators to create beautiful art. In keeping with its predecessors, this project is open source and can be remixed and added to. Feel free to fork and collaborate using GitHub.

Thanks for reading ✌