LoRA of Mihono Bourbon (umamusume), actually LoHa.
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. 👍
[2023-03-14 13:00 GMT] New version releases!
Many things change. Please use new version.
Recommend options
LoRA weight 0.7~0.8
Trigger words
base
mihono bourbon \(umamusume\)
Racing competition costume :
grey hairband, glowing hair ornament, detached sleeves, bare shoulders, clothing cutout, pink necktie, white leotard, highleg leotard, pleated skirt, grey skirt, lowleg skirt, miniskirt, white thighhighs, grey footwear
Halloween costume :
official alternate costume, ghost costume, halloween, mini top hat, black headwear, jack-o'-lantern on head, purple bow, hood up, white shirt, long sleeves, white jacket, chain
Valentine costume :
official alternate costume, valentine, white gloves, low-tied long hair, brown hairband, frilled hairband, hair ornament, jewelry, brooch, shoulder cutout, bare shoulders, red ascot, off-shoulder dress, sleeveless dress, wrist cuffs, brown dress, frills, waist apron, red footwear
Casual clothes :
casual, grey hairband, white hoodie, clothes writing, long sleeves, white jacket, open clothes, open jacket, yoga pants, blue pants, leggings, black socks, shoes, white footwear
Swimsuit :
grey hairband, indoors, water, covered navel, wet clothes, blue one-piece swimsuit, wet hair, poolside, competition school swimsuit, wet swimsuit, single vertical stripe
Tracen school uniform :
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
154 images of mihono bourbon
4x low quality 45 images
8x medium quality 55 images
16x high quality 54 images
=> 1 epoch = 1484 images
replaced character feature tags with mihono bourbon \(umamusume\)
1girl, animal ears, hair between eyes, horse ears, long hair, very long hair, brown hair, bangs, solo, tail, horse tail, purple eyes, blue eyes, tail through clothes, horse girl, ahoge
I think these tags are the essentials of bourbon.
resolution : 768x768 with aspect ratio bucketing.
other training settings are 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.
Description
LoHa version. extension needs.
FAQ
Comments (10)
Did you put loha in webui/models/lora folder directly?. Someone solve the problem in this way
@mht I did. mejiro mcqueen worked well this way. Maybe it's the difference between Locon and Loha. I will check the Nice Nature. Thank you for always making good files.
@mht Problem solved! After reinstalling the Locon Extension, it works fine. Sorry for bothering you.
"create LoCon for U-Net: 0 modules. enable LoCon for text encoder enable LoCon for U-Net shapes for 0 weights are converted. remove 'wrapped' from keys LoRA model mihonoBourbon_loha(5599e9e610f3) loaded: IncompatibleKeys(missingkeys=[], unexpected_keys=['lora_unet_input_blocks_1_1_proj_in.alpha',"
I got this argument despite using the kohakublueleaf links. Is there another extension for LoHa?
Did you see below messages in WebUI console when WebUI launch?
Additional Network extension not installed, Only hijack built-in lora
LoCon Extension hijack built-in lora successfully
@mht yes I did see the message
LoCon Extension hijack addnet extension successfully
LoCon Extension hijack built-in lora successfully
Did you use LoHa with Additional network extension? or LoRA notation <lora:name:weight> in prompts?
@mht I use additional network extension. I don't know why, but I keep getting message that I got the wrong namefile despite copypasting it on prompts.
@lisha Put loha file in your SD/models/lora and Use LoRA notation, <lora:name:weight>
Additional network extension does not support LoHa structure, not yet. Only you need is locon extension.
then, like another 'LoRA', you can use LoHa easily.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.



