Hi! Here is a LORA for SDXL. This Lora has been trained on about 1800 blip captioned images. I am going to re-visit the training with custom captioning and an improved dataset. For now, it works great as an asset creator for games, especially items, objects, and creatures (when well prompted). Still beta - so, no perfect as I wish (yet). *** KEYWORD *** In general - you don't need to use a keyword. But feel free to add "concept-sheet style ..." at the beginning of the prompt to strengthen the effect. Also "2d style anime" works wonders if style is still not there (in example if your prompt is describing raw materials like "lamp made from stone on stone with stone base and wooden decoration") ***======*** That's why I decided to release it earlier so people can use it for their own projects before I improve on it. This variant struggles with some "blur" effects sometimes, and overcomplexed characters due to some datasets. All dataset was based on AI generation. It works great paired with: <a target="_blank" rel="ugc" href="https://civarchive.com/models/120096/pixel-art-xl">https://civarchive.com/models/120096/pixel-art-xl</a> Enjoy!

Files with SHA256: 9fc7340ad7bc567035d42c17735f9def5ab871c107a97240df6f4d1b7fdcf7e7