Sign In

Twin Turbo (umamusume)

554
4.8k
34
Verified:
SafeTensor
Type
LyCORIS
Stats
4,783
Reviews
Published
Mar 23, 2023
Base Model
SD 1.5
Training
Steps: 4,010
Epochs: 10
Trigger Words
twin turbo \(umamusume\)
Training Images
Download
Hash
AutoV2
18238079F9
default creator card background decoration
Donator Badge
mht's Avatar
mht

LoHa of Double Jet (umamusume). Put LoHa file in your stable-diffusion-webui/models/lora folder and write LoRA notation in prompt to applying LoHa. <lora:your_loha_name:weight>

All of my works is free. If you'd like to support me, feel free to buy me a coffee. ๐Ÿ‘

Recommended options

  • LoRA weight around 0.7

  • Trigger words

    • Base

      • twin turbo \(umamusume\)

    • Racing competition costume :

      • stuffed bunny, hoodie, hood down, hooded coat, puffy long sleeves, multicolored clothes, multicolored jacket, drawstring, nail polish, black bodysuit, ankle boots, yellow footwear

      • (optional 1) pointing at self

      • (optional 2) aura, lightning, electricity

    • SSR Gut (BD):

      • casual, hairband, hair ribbon, brown ribbon, long sleeves, collared dress, brown dress, frilled dress, frills, white socks, brown footwear, mary janes

    • Casual

      • Due to lack of dataset. LoRA can not reproduce casual clothes of Duel Engine.

    • Tracen summer uniform

      • tracen school uniform, summer uniform, serafuku, puffy short sleeves, bowtie, horseshoe ornament, sailor collar, sailor shirt, frills, white skirt, purple shirt, miniskirt, zettai ryouiki, white thighhighs, brown footwear, shoes

Use a model derived from or mixed with animefull model.

  • Settings

    • Stable Diffusion WebUI by Automatic1111

    • 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

      • Upscaler

        • 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.

  • based on Animefull-pruned model

  • 176 images consist of

    • 4x low quality 61 images

    • 8x medium quality 60 images

    • 16x high quality 55 images

    • => 1 epoch = 1604 images

  • replaced character feature tags with twin turbo \(umamusume\)

    • Relation with Uma Musume

      • 1girl, solo, horse girl, horse tail, horse ears, animal ears, tail through clothes,

    • Hair

      • long hair, very long hair, twintails, hair between eyes, purple hair, blue hair, aqua hair, two-tone hair, multicolored hair, bangs, crossed bangs, ahoge, green hair, gradient hair, hair ribbon, striped ribbon, white ribbon, ribbon, black ribbon, ear ribbon, hair bow,

    • Eye

      • @_@, heterochromia, pink eyes, purple eyes, aqua eyes, red eyes, blue eyes,

    • Face

      • sharp teeth,

    • Unnecessary

      • stuffed toy, stuffed animal, simple background,

  • I think these tags are

    • the essentials of Twin Turbo

    • or unnecessary for training because of duplication.

    • twin turbo \(umamusume\) contains above tags.

  • 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.