I can guide you with the best practices, tips, and tricks to achieve your goal in your Stable Diffusion training. Take a look at my Patreon, subscriptions for just $5. Exclusive content. Patreon. :)

About:

The purpose of this LoRA is to (try to) replicate a woman with her shirt lifted.

It was trained in multires (512, 768, 1024), using simple captions and with 55 high-resolution images.

The demo images of the model aren't edited, but they are cherry-picked and generated using hires-fix (0.25 ~ 0.45 denoising).

Suggested Settings for Inference:

Model: cyberrealistic_v40 (but works with 97,2% of models)

Positive: a woman, head and chest only, realistic body structure, slender and muscular build, strong legs, sitting, very curvy, sizes and colors, epic proportions, 2000s photo, sculptured, outdoor scene <lora:Upshirt_v1_SD15_by_IsnAI:1>

Negative: snap-fingers:2, inaccurate-number-of-fingers:2, inaccurate-number-of-limbs:2, crooked-fingers:2, crooked-limbs:2, open mouth

Sampler: DPM++ 2M Karras

Steps: 32

CFG: 7

Resolution: 512x768

Upscaler: 4x_NMKD-Siax_200k

Note: You can get great image variations if you tweak resolution a bit.

Triggers:

Main: a woman is posing

Angle: head and chest only, full scene shot

Tags Used on Training: standing, sitting, kneeling, squatting, indoor scene, outdoor scene, profile, from behind, shot from below, squatting