Sign In

bikabaka / ビカパカ style | NAI style

542
3.3k
209.8k
237
Updated: Jan 25, 2025
styleanimewomancutegirlslora
Verified:
SafeTensor
Type
LoRA
Stats
2,530
129,196
Reviews
Published
Nov 16, 2024
Base Model
Illustrious
Training
Steps: 21,093
Epochs: 10
Usage Tips
Clip Skip: 2
Strength: 0.8
Hash
AutoV2
AE33EA04C1
default creator card background decoration
wwHEY's Avatar
wwHEY

Please support the original creators.

This model was not created to harm or otherwise inconvenience the original creators nor Civitai or its users. It can be removed upon request.

Illustrious -> Make sure to include "nsfw, r18" in your negatives to generate sfw images.

24-12 Fixed file name, either change your local copy to 'bikabaka.safetensors' or re-download the model. Still broken !

Model based on the works of Bikabaka or ビカパカ on Pixiv. Give them a follow if you enjoy this style; be warned that their content may not appeal to everyone.

This style also relies on the works of anti-freez3 (天凉多喝防冻液) on Pixiv. They own the original art style this model is based on. As with the previous creator, support them however you can. Same content warning applies.

Model trained with IllustriousXL. Most other checkpoints based on Illustrious will be able to achieve a similar result.

Illustrious model

A typical prompt for this style would be formatted as follows:

[Rest of the prompt], tianliang_duohe_fangdongye:[STR], <lora:bikabaka:[STR]>

Recommended settings (external gen-comfyui):

  • lora weight: 0.7 and 0.8 respectively

  • sampler: euler a

  • scheduler: sgm_uniform/normal

  • steps: 20-40

  • CFG: 5.0-7.0

Recommended settings (civitai generator):

  • lora weight: 0.5 and 1.0 respectively

  • sampler: euler a

  • steps: 30-40

  • CFG: 3.0

  • denoise: 0.6

  • hires fix

All images in the showcase use Ultimate SD Upscale 4xAnimesharp for upscale.

PonyV6 model

Want a simpler style with more compatibility with other loras ? Use 1.1 !

Want a similar style to illustrious ? Use 1.0 !

IMO, 1.1 > 1.0

Use the appropriate score tag, as well as the lora:

positive -> contains optional tags to achieve a specific style

score_9, score_8_up, score_7_up, [Rest of the prompt], <lora:bikabaka_pony:[STR]>

negative -> use these to adjust the output and remove unwanted elements

score_6, score_5, score_4, worst quality, 

Recommended settings (external gen-comfyui):

  • lora weight: 0.8-1.0

  • sampler: euler a, dpmpp 2m

  • scheduler: sgm_uniform/normal/karras

  • steps: 20-40

  • CFG: 3.0 (softer, bloom style), 5.0 (solid lines), 8.0 (strong lines)

  • upscale recommended

[Testing with civitai generator in progress]