VoluptuousMix v1.0

This is the first checkpoint I've ever published. If I had known that merged models were so easy to make, I'd have done it long ago. But here we are.

This model begins with an all-purpose SD1.5 model and adds a few other models that emphasize voluptuous, busty women. Those additional models, when used on their own, tend to be limited in background and face variety, so I created this merge to add in a pile of quality faces, scenes, clothing, and camera angles.

The images may not quite live up to "voluptuous" without any keyword guidance, but the model responds well and strongly when you suggest things like "large breasts", "busty", and so on. Using more than one of these terms in a prompt, especially if you boost the weight (like "(huge breasts:1.2)" ) will bring out results in a hurry. In my experiments creating test images from this checkpoint, it worked just about as well as most Loras I've used that aim for the same thing.

Speaking of Loras, the checkpoint works well with the ones I've tried. It works in a variety of image sizes and also responds very well to SDXL style prompts. I've found that the SDXL prompts are a great replacement for copy/pasting "ultra-super-duper-masterpiece-realistic whatever blah blah" onto every image I make.

If this works well, I may try and do a similar kind of thing with SDXL, which is where I spend most of my rendering time anyway.

The sample images I'm including here are all created without the help of controlNet, aDetailer, or upscaling. They are provided here straight out of automatic1111 at their original resolution. All the prompts are included so feel free to use these as a basis for your own work. Or don't! Whichever.

For my test, I got the best results with CFG between 2 and 4, steps usually around 20-30 but sometimes higher, and often I didn't need any negative prompts at all. I like to render with DPM++ 3M SDE and SGM Uniform but as always that's a matter of personal taste.

If you'd like to play with SDXL styles, even in your SD1.5 images, you can read about them here: https://stable-diffusion-art.com/sdxl-styles/

You can copy/paste the prompt information straight from that site in order to try things out, and if you like the results you can find extensions that will apply the styles to your prompts for you. I use one in automatic1111 and it works really well.

I hope you have fun with the new model!