As someone who has gained tremendously from the stable diffusion (SD) community, I am thrilled to have the opportunity to give back by sharing my work. I've fine-tuned an SD model on over 15,000 Midjourney v4 and v5 images, resulting in a powerful tool that captures the unique essence of MidJourney with capability of generating NSFW.
You can run this model on https://randomseed.co/model/43 with GPU acceleration.
By supporting me on Ko-fi (https://ko-fi.com/michelangelofussion), you're not only empowering me to maintain and improve the current model, but also enabling me to expand my research and train even more advanced models on 100k and 1M images. Your valuable contributions will help cover server costs and propel us further on this incredible journey.
My model, trained with offset noise, effectively grasps the MidJourney aesthetic, especially when it comes to generating art-like images. While prompting photorealistic content is possible, it requires more effort due to the model's complexity. The model can generate images in various sizes, but 512x512 has proven to produce the fewest artifacts.
This model is freely available for use under any circumstances, except when utilized as part of merges offered on paid platforms.