Type | |
Stats | 6,285 |
Reviews | (1,133) |
Published | Feb 26, 2023 |
Base Model | |
Training | Steps: 4,300 Epochs: 20 |
Trigger Words | pmini style miniature |
Hash | AutoV2 5BF44127B9 |
A lora model dedicates to generate painted miniature figurine. The lora model is trained with SD1.5 base model. The example images are generated with Clarity and Protogen x5.3 (Photorealism) ckpt models. It should work fine with other fine tune SD1.5 originated ckpt models.
User suggestions:
The following setup gives me satisfied outputs.
trigger words: painted miniature, pmini style, on a base (optional: to add a plate under miniatures' feet)
lora weight: 0.6-0.8
Sampler: Euler a for 100 steps
W*H: 768*1024
CFG scale: 5-7
Negative emb. (optional): bad_prompt
Hires. fix: R-ESRGAN General WDN 4xV3, steps 20, denoising strength 0.4
Prompt example 1: miniatures in themed background
positive prompt: a painted miniature of a male human thief on a base, in a lost ruin, pmini style, character design, studio lighting, (extremely detailed 8k wallpaper), Intricate, High Detail figurine, Sharp focus, dramatic, photorealistic art <lora:Painted_Miniature_i_v3.0:0.6>
negative prompt: headpiece, (bad_prompt:0.8), cartoon, (bad art:1.2), (deformed:1.2),(close up:1.2),(b&w:1.2),blurry, (duplicate:1.3), (mutilated:1.2), (out of frame:1.2),blurry, (disfigured:1.3), closed eyes
Prompt example 2: high detailed miniatures in plain background
positive prompt: pmini style, painted miniature of male human thief, character design, Intricate, High Detail, Sharp focus, dramatic, photorealistic art <lora:Painted_Miniature_i_v3.0:0.6>
negative prompt: headpieces, cartoon, (bad art:1.2), (b&w:1.2),blurry, closed eyes, nsfw