Hyper JK Bosom V03 LoRA

Harder, better, faster, faster, faster, faster👤

This Lora allows the Bosom V03 checkpoint to work like Hyper SD 1.5. A test that ended up bringing great results.

Based on ByteDance models, but the precision and weights have been fine-tuned by me.

🖼️ Gallery

Gen Parameters:

• Sampler: DPM++ SDE Karras (practically only this one works well)

• Sampling steps : 5 to 7

• CFG: 4 to 5

Hires Parameters:

• Sampling steps : 3 or 4

• Same sampler and same scheduler.

• Upscale by 1.5 to 2

Limitations:

This Lora has some problems with the side viewpoint.

OBS: I'm already starting to collect information and adapt the dataset for SD XL training.