TRAINED ON ANIME-FINAL-PRUNED

What's New?

1. Better costumes

2. Fixed wrong colours on some costumes

3. Fixed eyes

4. Give more details to the face

5. More data sets to get more accurate looks

6. This version is using 19-epochs to train

see my fanbox for more explanations

Here are the trigger words of each costumes:

各衣装のトリガーワードは以下の通りです。

1) Default >>> Hiei, a woman in a white_HI_uniform, white cape, white_Hhorns, HIE_hairaccessories on her head, holding a katana sword

2) Rigging >>> Hiei, a woman in a white_HI_uniform, white cape, white_Hhorns, HIE_hairaccessories on her head, holding a katana sword, rigging_machine on her back, turrets_equipped_on_rigging_machine

3) Beauty of the White Sands >>> Hiei, a woman sitting on a beach next to a banana boat, wearing white_BWS_dress, ponytail hair, wearing BWS_hairaccessories on her head, sand_castle_turrets in the background

NEW TRIGGER FOR MOONLIT COSTUME

Moonlit Cruise >>> Hiei, a woman in a moonlit_cruise_costume sitting on a bed, turret_pillow on her sides, white_Hhorns, moonlit_hairaccessories on her head

NEW TRIGGER FOR FRAGNANT RED PLUM

Fragrant Red Plum >>> Hiei, a woman sitting on a desk with a chinese_partition in the background, purple flower embroiled, wearing black CNYdress, holding a fan, white_Hhorns, FRP_hairaccessories on her head

DON'T GO OVER 1 LORA STRENGTH AS THIS MODEL IS VERY OVERFITTED WITH DATA

RECOMMENDED LORA STRENGTH:

10 EPOCH >>> 0.4-0.8

13 EPOCH >>> 0.4-0.8

19 EPOCH >>> 0.4-0.8

SWEET SPOT >>> 0.7-0.8

このモデルはデータで非常にオーバーフィットしているため、1 LORA STRENGTH を超えないでください

推奨されるロラの強さ:

10 エポック >>> 0.4-0.8

13 エポック >>> 0.4-0.8

19 エポック >>> 0.4-0.8

スイートスポット >>> 0.7-0.8