
Atelier Totori - Totooria Helmold トトゥーリア・ヘルモルト LoRA
Trained mostly based on official artworks.
Recommended CFG: 5-6
Recommended weight: 0.8-1 (depends on your personal settings)
Recommended prompt: blue sleeves, detached sleeves, skirt, dress, bare shoulders, blush, collarbone, strapless, hat, frills, flat chest, bow, back bow,
NOTE: if you do not want to get ribbon or bad back bow generation results, removing bow, back bow from your prompt can help you.
I used to train 7 version of Totori LoRA, this LoRA version is the best one.
The reason her LoRA is so difficult to produce is largely due to her clothes. With a limited training sample, you'll have a hard time finding quality clothing details to use as a reference. So I had to spend more time on tag editing. I tried generating it several times and it has a relatively good performance at 768*1024 resolution.
描述:
First version
Existing possible problems:
-
blurry hat/hair ornament
-
slightly incorrect dress (not often happen)
-
back bow, bow, ribbon bad generation results
The refined version is on my schedule. I probably will redraw some sample images by myself.
训练词语: chartotori
名称: totoriv1f.safetensors
大小 (KB): 36980
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success