This LoRA lets you edit the bulge in the crotch in clothing. Since MOSS-Image is not a true edit model, it requires the use of a custom node from Ostris for the LoRA to work with it, therefore, this cannot be used with the online generator.

Custom node info:
https://github.com/ostris/ComfyUI-MOSS-Image-Ostris-Edit

He has an example workflow included with the custom node package.

This LoRA has been trained against the Base model, which works with the SFT model as well. I also trained a version against the SFT version and that is in the Optional Files section. For my workflow, it seems that the SFT model with my SFT trained LoRA is giving the best results. You can be really specific about what you want the bulge to look like in the prompt.

Example Workflow
For no turbo LoRA, I'm finding 30 steps, cfg 3, euler/simple is producing the best results. If you want to use the 8 step turbo LoRAs you'll see it is so much faster, and quality is good with 8 steps, cfg 3, er_sde/bong tangent or sgm uniform. I've posted my workflow since it makes it easy to switch between Base and SFT model, and with or without the turbo Lora. If you want to use the base model, toggle your loras for that manually in the power lora node, and toggle Use SFT not base model to false and it will use the base model. You will most likely need to update model paths inside that Load Models and Process Image node. I'm using Pixaroma nodes and other common custom nodes.

NOTE: I did try out the new Moss-Image Edit node - but it definitely uses different settings that this workflow and felt this one gave me more consistent results. DM me if interested and I can share what I did on that front.