abbuibuibui / MoehaFujiwara-Lora

huggingface.co
Total runs: 277
24-hour runs: 10
7-day runs: 108
30-day runs: 277
Model's Last Updated: September 30 2026

Introduction of MoehaFujiwara-Lora

Model Details of MoehaFujiwara-Lora

Moeha Fujiwara (Kaguya-sama) — Illustrious SDXL Character LoRA

中文说明 | English (current)

An unofficial character LoRA for Moeha Fujiwara (四条真妃 / also known in fandom databases as Maki Shijo) from Kaguya-sama: Love is War , trained on waiIllustriousSDXL v17 ( waiIllustriousSDXL_v170 ). This repository publishes both training-run epoch weights. Comparative evaluation recommends R2E5 ( maki_shijo_rE4S582-000005.safetensors ) as the default at strength 0.8 .

Trigger words stay maki_shijo / maki_casual (training captions). The public model title and repo id use Moeha Fujiwara .

Training images and captions: abbuibuibui/MoehaFujiwara-Dataset .

This is a fan-made derivative, not an official product. Rights to the character and source work remain with Aka Akasaka / Shueisha.

Files

Step checkpoints ( *-step* ) and optimizer state dirs are not published.

Label File Role
R1E1 maki_shijo-000001.safetensors Alternate (first training run).
R1E2 maki_shijo-000002.safetensors Alternate.
R1E3 maki_shijo-000003.safetensors Screened into early matrix; softer face / heavier blush.
R1E4 maki_shijo-000004.safetensors Best of run 1; weaker full-body / more leak than R2E5.
R1Final maki_shijo.safetensors Unnumbered final from run 1 (SHA differs from R1E1–E4; kept as-is).
R2E4 maki_shijo_rE4S582-000004.safetensors Resume run from E4 @ step 582; solid profile/night.
R2E5 maki_shijo_rE4S582-000005.safetensors Recommended default @ 0.8.
R2E6 maki_shijo_rE4S582-000006.safetensors Close to E5; more hair-ornament lock-in at high strength.
R2Final maki_shijo_rE4S582.safetensors Unnumbered final of resume run (SHA differs from R2E6; kept as-is).

R2E5 SHA-256: A7850792C5888834566A4D5C629A221218704AAB48D6A46547ED945FF967C61D

Showcase

Direct R2E5 @ 0.8 outputs, no retouching. Euler a, 28 steps, CFG 4.5, CLIP skip 2, 832×1216 ; Hires / ADetailer off.

Portrait Casual full body
Portrait Casual
Night street No outfit token
Night No outfit
Trigger Words and Settings

Captions used keep_tokens = 2 , shuffle_caption = true . Dataset: 17 image/caption pairs, all maki_casual . num_repeats = 10 on the resume run. Network dim 32 / alpha 16, AdamW 8-bit, UNet + both text encoders, bf16, SDPA, 1024 buckets. Training log does not print LR, so this card does not invent one.

Role Token Notes
Identity maki_shijo Light pink hair, blue eyes, short twintails, green hair bows
Outfit maki_casual Only outfit token in the training set

Recommended: waiIllustriousSDXL v17 / Illustrious family; strength 0.8 (0.6 already recognizable; 1.0 turns the face / locks accessories). Euler a, 28, CFG 4.5, CLIP skip 2, 832×1216 .

Example:

<lora:maki_shijo_rE4S582-000005:0.8>,
masterpiece, best quality, very aesthetic, absurdres, anime screencap,
1girl, solo, maki_shijo, maki_casual,
light pink hair, blue eyes, short twintails, green hair bows,
portrait, looking at viewer

Negative (baseline + color-bleed guards):

lowres, bad, bad anatomy, bad hands, text, error, missing fingers, extra digit,
fewer digits, cropped, worst quality, low quality, low score, bad score,
average score, signature, watermark, username, blurry, extra person, two girls,
multiple girls, blonde hair, dark green hair, green eyes, purple eyes,
red hair, black hair
Why R2E5

Round 1 screened both runs at 0.8. Round 3 matrix used candidates R1E3 / R2E4 / R2E5 across 10 scenes × 0.4/0.6/0.8/1.0. R2E5 was the most balanced for face, casual full-body, night, and leak control. See evaluation/横向测评记录.md .

Public Horizontal Evaluation

Sheets: columns = candidates (R1E3 / R2E4 / R2E5), rows = weight 0.4–1.0. Original outputs only.

Identity / angle
Front Three-quarter
Face 3/4

Profile

Full body / outfit
Casual full School (untrained token)
Casual School

Sitting

Scene / leak / trigger
Simple BG Complex night
Simple Night
No outfit words No trigger
No outfit No trigger
Training
  • Data: 17 curated stills + matching .txt . Captions start with maki_shijo, maki_casual . Published as abbuibuibui/MoehaFujiwara-Dataset .
  • Two training sessions share one output\ : run 1 ( maki_shijo-* ) and resume run 2 ( maki_shijo_rE4S582-* from E4 / step 582).
  • Network: LoRA rank 32 / alpha 16, UNet + both text encoders.
  • Optimizer: AdamW 8-bit. LR not printed in logs.
  • Precision: bf16, SDPA, 1024 buckets, batch 1, latents cached. Flip / color aug off.
Related repositories

Civitai: recommended weight only (R2E5) — link filled after publish.

Compatibility and Limits

Scored on waiIllustriousSDXL v170, Euler a, 832×1216 . Character LoRA only. Only maki_casual was trained as an outfit token; generic school uniform is a generalization probe. At high strength accessories may lock in.

License
  • Character / source rights © Aka Akasaka / Shueisha / production committees.
  • Weights under CreativeML OpenRAIL-M .
  • Fan non-commercial use intended; follow the base model license for generations.
  • Do not use for unlawful, deceptive, harassing, or exploitative purposes.

Runs of abbuibuibui MoehaFujiwara-Lora on huggingface.co

277
Total runs
10
24-hour runs
14
3-day runs
108
7-day runs
277
30-day runs

More Information About MoehaFujiwara-Lora huggingface.co Model

More MoehaFujiwara-Lora license Visit here:

https://choosealicense.com/licenses/creativeml-openrail-m

MoehaFujiwara-Lora huggingface.co

MoehaFujiwara-Lora huggingface.co is an AI model on huggingface.co that provides MoehaFujiwara-Lora's model effect (), which can be used instantly with this abbuibuibui MoehaFujiwara-Lora model. huggingface.co supports a free trial of the MoehaFujiwara-Lora model, and also provides paid use of the MoehaFujiwara-Lora. Support call MoehaFujiwara-Lora model through api, including Node.js, Python, http.

MoehaFujiwara-Lora huggingface.co Url

https://huggingface.co/abbuibuibui/MoehaFujiwara-Lora

abbuibuibui MoehaFujiwara-Lora online free

MoehaFujiwara-Lora huggingface.co is an online trial and call api platform, which integrates MoehaFujiwara-Lora's modeling effects, including api services, and provides a free online trial of MoehaFujiwara-Lora, you can try MoehaFujiwara-Lora online for free by clicking the link below.

abbuibuibui MoehaFujiwara-Lora online free url in huggingface.co:

https://huggingface.co/abbuibuibui/MoehaFujiwara-Lora

MoehaFujiwara-Lora install

MoehaFujiwara-Lora is an open source model from GitHub that offers a free installation service, and any user can find MoehaFujiwara-Lora on GitHub to install. At the same time, huggingface.co provides the effect of MoehaFujiwara-Lora install, users can directly use MoehaFujiwara-Lora installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

MoehaFujiwara-Lora install url in huggingface.co:

https://huggingface.co/abbuibuibui/MoehaFujiwara-Lora

Url of MoehaFujiwara-Lora

MoehaFujiwara-Lora huggingface.co Url

Provider of MoehaFujiwara-Lora huggingface.co

abbuibuibui
ORGANIZATIONS

Other API from abbuibuibui