Visit My Site For Models I Can't Post On CivitAI: https://qtgurl.com/

To get the results you see in the images above, these are the settings I used.

Version 2: Anima

This lora was trained on Anima 1.0 using 54 images of various backgrounds with no humans. The dataset is a mix of similar and different images compared to V1. I’d say about 60% of the images are outdoor with 40% of them being indoor. Most of the material used are from city scenes, but there are also images from open fields, forest and beaches.

I can’t tell you the amount of times I tried to make this lora again. SDXL was just awful with backgrounds. Anima is much better but my old training settings were negatively impacting the results. The settings I’m using now have allowed me to produce something that is much closer to the source images.

Version 1: SDXL

This lora was trained on AutismMix using roughly 60 images selected for their vibrant and detailed scenery. My goal was to create a lora that could be used in conjunction with character loras to enhance their backgrounds. What this lora does best is cities, nature, beaches and specific details such as random items in a scene. Most of the example images are the lora alone, a few of them used my other lora MeMaxXL to showcase how it enhances the background. More than anything, this lora increases the dynamic range of color compared to base Pony/AutismMix.

If you compare it to my other scenery lora, CoMix BGXL the details and colors are much more prominent and defined. You can try them both and determine which you like best.