Quick model of Kirsten Dunst based on a dataset that was shared with me to create a LoRA.
Turned out good with no modifications to the source data and caption.
Utilise the kohya-ss/sd-webui-additional-networks (github.com) in auto1111 to load the LoRA model. You can also blend it with other LoRA models for a diverse range of poses.
Can be combined with Kirsten Dunst | Stable Diffusion TextualInversion | Civitai at 50/50 ration for great variations
Description
Created from mixed aspect ratio images rather than square ones like in v1. Produce results where it is easier to get full body shot since the source images contained full body shot of the subject.
For good results:
Use LoRA strength of 0.75
Use DPM++ SDE Karras sampling
Use hires fix with denoising strength of 0.5