A general purpose detailer, trained on images captioned both with Florence 2 Large rich captions and also WD-EVA02-Large-Tagger-V3 Booru tags. No particular trigger word. Recommended strength is around 0.6 to 0.8 or so, but feel free to adjust the strength as you see fit for any particular image.
Note that you CAN combine this Lora with my Zoot Fluxxer XL Lora if you wish, as both Loras were trained on completely different datasets.
Description
Initial Version.
FAQ
Comments (4)
somehow it workes really good with negative -0.7 for me : ). with IPadapter. The pictures are cleaner but still very detailed.
Kind of random but when it works it works I guess. Thanks for sharing!
Need to do a bit more testing
The last couple images I generated onsite have the "before" with just base SDXL too if you click into the post.
what kind of detail you talk about here ?
Basically any, I had a lot of different stuff in the dataset. It can improve prompt adherence too cause all the images are captioned both with detailed captions and also booru tags.