Type | |
Stats | 5,020 86,559 |
Reviews | |
Published | May 28, 2023 |
Base Model | |
Hash | AutoV2 C5717CC80A |
SDXL version of model trained with original SDXL checkpoint, tested with RealVisXL V2.0
Trigger words: footpose, barefoot, male feet, sitting, soles, toes, foot focus, solo
Lora weight: 0.85 - 1.0
CFG scale - 5-7
Some negatives: bad, worst, floating,(shoes, socks, nude:2), standing
0.5v is an early version, but it seems to work. I hope to find a way to train a better one.
SD1.5
↓
Lora model for male pose with bare feet (trained with SD1.5), focused on realistic style.
Recommended settings v2:
Lora weight: 0.8 - 1.0
Image dimentions: 768x786 and other square format is preferred
prompt example: a photo of handsome man showing his feet, foot_focused, wearing black boxer-shorts, green field photo, sitting in wild grass field, feet up, barefoot, toes, soles, long legs, bent knees, spread legs pose, composition is awesome <lora:footpose-v2:0.8>
CFG scale: ~7
negative prompt example: (low quality, worst quality), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), legs crossed
Tested with Adobi's Real Flexible Mix and Something. Should be more consistent then v1.
Recommended settings v1:
Lora weight - 1.1 ±0.1
prompt: (feet, soles, toes, feet up, man, barefoot, sitting,view from below)
CFG scale - 7
negative prompt: (low quality, worst quality:1.4), (inaccurate limb:1.2), bad composition, extra digit, fewer digits
Tested with hyperV1_v10 model. This Lora is far from perfect, but it can give some good results.