LoHa of Wonder Acute (umamusume). Please install extension on WebUI.
All of my works is free. If you'd like to support me, feel free to buy me a coffee. 👍
Recommend options
LoRA weight 0.7~0.9
Trigger words
base
wonder acute \(umamusume\)
Racing competition costume :
choker, black gloves, fingerless gloves, detached sleeves, puffy long sleeves, frilled sleeves, pink dress, frills, pink jacket, striped skirt, black pantyhose, brown footwear, knee boots, high heel boots
Casual clothes :
casual, white shirt, necklace, long sleeves, white shirt, white sweater, belt, green skirt, plaid skirt, long skirt, black pantyhose, shoes, sneakers,
(optional) cup, holding cup, green tea
Gaze on me :
(white shirt:1.2), off-shoulder shirt, tied shirt, short sleeves, midriff, scrunchie, pink shorts, swim trunk
(optional) on stage, concert
Tracen school uniform(summer) :
tracen school uniform, purple shirt, pleated skirt, puffy short sleeves, white skirt, puffy sleeves, summer uniform, frilled skirt, sailor collar, sailor shirt, miniskirt, frills, white thighhighs
Use a model derived from or mixed with animefull model.
Settings
DPM++ Series(SDE Karras, 2M Karras, etc.)
about 20 steps, CFG scale 3.5~6.5
I recommend kl-f8-anime2.vae
CLIP skip = 1
Use Hires.fix to get higher quality image
Up-scaler
Latent, Latent (nearest-exact), Latent (bicubic antialiased) or other Latent series.
Denoising strength 0.50~0.65
LoRA training info
trained on sd-scripts by kohya_ss.
Thank you all a lot!
based on Animefull-pruned model
54 images of wonder acute
4x low quality 20 images
8x medium quality 15 images
16x high quality 19 images
=> 1 epoch = 504 images
replaced character feature tags with wonder acute \(umamusume\)
1girl, horse tail, tail, grey hair, horse girl, brown hair, bangs, multicolored hair, green eyes, blue eyes, solo, streaked hair, simple background, very long hair, two-tone hair, animal ears, horse ears, long hair, white hair, hair between eyes, thick eyebrows
I think these tags are
the essentials of wonder acute
or unnecessary for training because of duplication.
resolution : 768x768 with aspect ratio bucketing.
other training settings is included in training_data.zip or metadata in LoRA file.
All uploaded images are generated by AOM3A1
And, special thanks to my best friend. DH.