>>2762

As far as I know there shouldn't be too much difference between style- and concept- training. Instead of using a triggerword for your style (or leaving the Lora completely triggerless) you'd have to specify a triggerword for the action ("gutstab" perhaps?) and the kind of weapon you end up using. In case you didn't have to worry about tagging at all so far, this guide may be helpfull: https://civitai.com/models/351583/sdxl-pony-fast-training-guide ...also some of the settings should hopefully apply to Onetrainer, too. But again, all I've said is laymans speculation at this point. I may have curated/created the images for the pussystab_v1 Lora's dataset, but it was Entropy who was so kind to tagg and train the Lora with it. He also gave me the link to that guide. But I'm still a few weeks away from having enough data for me to try to train a _v2 (multiple _v2's) on my own. So hopefully someone with greater experience in creating Loras could correct what I've said so far or add relevant info I've missed.