BulletHits-LoRA-V1 for SDXL

A LoRA model as an extension for Stable Diffusion XL (SDXL)

The LoRA helps to create hits from bullets with large or small fountains of blood. The effect should be dramatic and exaggerated, like in comics or corresponding genre films. It has tags for head and torso hits, each with 3 different strengths. It is also possible to hit multiple times, and the package also includes two special LoRAs for this purpose.

The LoRA has also been trained for the corresponding poses of the characters and as a supplement it is possible to generate a sandbag wall with barbed wire for typical war execution scenarios.

As usual with AI, the LoRA does not always do what you want and especially with multiple hits not every image is good. Inpainting is especially helpful for such scenarios.

The promo images are all directly from AI and not post-processed, they were created in Fooocus, mainly with the standard and the Juggernaut8 checkpoint

It was trained with about 2000 images and 23 tags and can be used with many styles.

The package contains 4 LoRA versions:

2 "BulletHitsXL" and 2 "BulletHitsMultiXL" LoRAs each

.There is a version 1 and a version 1.1 of both.Version 1 is actually the most suitable and should be used mainly. Version 1.1 is a more trained version that may be more suitable in special cases with many other LoRAs or other models or tags.

Of course you need Stable Diffusion XL and one of the many user interfaces to use it. A small description is included.

The model does not always produce perfect images. But it does produce a fair share of great images. And since the image generation in SDXL does not take long, you can live with the waste or use inpainting and other techniques to repair the imperfections. It also seems to depend heavily on the theme and style whether the model can handle it or not. Therefore, 4 differently trained versions are included. It always depends on the scene, models used, trigger words, etc. which LoRA is most suitable. So you can try and choose for yourself.

In addition, the readme file contains 6 example prompts, which were also used to create the promo images, these are intended as a starting point.