This model was merged for the purpose of uploading a LoRA.
Description
FAQ
Comments (1)
Didnt expect that. Could use bf16 as I need to make quants for myself usually.
Qwen Image 2.1 is also good target, especially due size and its fairly smart. Just LoRA training is harder (well not actual training, but figuring out how to make result sharp).
https://civitai.red/articles/35836/lora-surgery-on-qwen-image-21-sharper-skin-without-retraining
Just in case you ever wanted to do bolero LoRA for it, might be helpful as that "soft results" plague not just his specific LoRA, but almost any LoRA I tried so far.






