Please don’t rely only on the LoRA title to understand what it does. The preview images are the real representation of the results you should expect.
My LoRAs are trained on small, focused datasets, so they naturally stay close to the original concept. If a dataset contains green bedrooms, livestream UI, or Smurf-style houses, the LoRA will strongly lean toward those elements. Trying to turn it into something completely unrelated usually will not work well.
Please use the preview images as your main reference and keep the dataset concept in mind while prompting.
If you like one of my LoRAs but want different colors, atmosphere, or no livestream UI, feel free to ask. I’m always happy to remake or adapt a concept into a different version.
Description
FAQ
Comments (1)
Looks like an interesting lora. However, I imagine the bulk of the people using this lora, would want to generate images involving the rooms to the sides, rather than the hallway itself.
That said; as someone that's attempted to create an area lora (Dundas Square, Toronto), and failed pretty hard, I can understand why you'd want to make it more focused on an ultra specific location.
Though, I'm not sure which training program you use, or if you just use some cloud thing, but in OneTrainer, you can add multiple different 'concepts' to the lora for training, and so, splitting up the room/hallway into separate concepts, could enable you to create a more versatile lora which is capable of greater exploration of the area.
Just being sure to add distinct tags for each separate area.
Though, I haven't attempted this strategy yet in my own loras, it does seem logically sound. Oh, and it may also help further to ensure a balance in data for each concept, depending on the size of the concept in question. For instance, if the room is very small, and lacks detail in contrast to the hallway, then less data may be better so as not to over-train a simple concept.