Type | |
Stats | 695 4,378 |
Reviews | (99) |
Published | Aug 27, 2023 |
Base Model | |
Training | Steps: 4,000 Epochs: 25 |
Usage Tips | Clip Skip: 1 |
Trigger Words | lasgun |
Hash | AutoV2 23A1371939 |
A LoRA to allow you to incorporate Lasguns from Warhammer 40k into your AI art. It also biases armor pretty heavily towards Cadian/Kasrkin armor unless told otherwise, but it will happily do other outfits.
I recommended weight 0.9-1.0, depending on model. I've been getting reports that on other models, lowering the weight to 0.7-0.6 works better. Experiment!
On mine, RoXL, it works best at 0.9, while on BriXL it works better at 1.0. Somewhere in this rough range should work. I also strongly recommend terms like "rifle" and "holding gun" to help guide SDXL.
The success rate is reasonably good. In my testing about 50% of the lasguns were passable, while maybe 1/8- 1/10 were excellent. For best results, though, slight inpainting is helpful, though the example images are not inpainted.
Finally, training models takes a lot of work, and to train SDXL models I need to pay for time on higher end GPUs, so if you'd like to support me experimenting and making models, feel free to toss me a dollar or two on ko-fi.