Hi everyone, I'm having trouble creating the dataset. Could you briefly explain how you do it? Thanks to Icekiup, I've made some progress, but I'm not getting consistent training. Whether running a runpod or locally, what parameters are used for training? Specifically, what about the dataset in terms of identity, body type, and number of images? And if we want to create a commercial version of LoRa for FanVue or other sites with this type of content, we need consistency and coherence in anatomy, as well as realism. I've tried SDXL, which is more or less manageable for this type of content, but I'm not getting good results because ZImage or Flux Klein, among others, make it difficult to create NSFW content for a character. I appreciate your answers, guides, and opinions.🙏👍

Anyone here managed to keep a consistent background in ComfyUI while changing poses/outfits? What workflow are you using?