Aesthetic detailed
A LoRA that aims to enhance the light, composition and background of images
Base model: Chenkin noob xl v0.5
Trigger: aesthetic detailed
Strength: 1.0 or lower
Apply model only.
Latest release:
v0.5_stable_aggressive:
Largely increase effect on composition, its might be the latest version for long time.
Special: As civitai recently can't count downloads, leave a like to let me know the response of this LoRA in community.
Versions:
v0.5_aggressive:
dim: high(64)
resolution: 1216x1216
dataset: 1009 images
training time: 6 GPU hours on RTX5090D
v0.4_experimental:
dim: medium(48)
resolution: 1216x1216
dataset: 1009 images
training time: 3 GPU hours on RTX5090D
v0.3:
dim: Low(32)
resolution: 1216x1216
dataset:1421 images
training time: 3 GPU hours on RTX5090D
testing time: 28 GPU hours on RTX5090D
v0.2_mini:
dim: Low(32)
resolution: 1152x1152
dataset:1596 images
training time: 8 GPU hours on RTX5090D
v0.1_highdim:
dim: Xhigh(96)
resolution: 1152x1152
dataset: 1983 images
training time: 6 GPU hours on RTX5090D
v0.1:
dim: Low(32)
resolution: 1216x1216,1408x892
dataset: 1983 images
training time: 5 GPU hours on RTX5090D
Note:
watermark and copyright problem will be fixed in v1.
The LoRA is in developing version, further training in progress, welcome to give suggestions.
Description
v0.1_large_preview:
Trained on larger LoRA network(dim = 96)
Trigger: aesthetic detailed
Base model: Chenkin noob xl v0.5
resolution: 1152x1152
dataset: 1983 images
training time: 6 GPU hours on RTX5090D
suggest to use 1.0 or lower strength.
Note that this is only the preview version, further training in progress