abbuibuibui/emilia-Dataset

Emilia (Re:Zero) — Illustrious SDXL Character LoRA Training Dataset

中文说明 | English (current)

This is the actual training subset used to train abbuibuibui/emilia-Lora, an unofficial Emilia character LoRA on waiIllustriousSDXL v17. The files here are a copy of the directory named in dataset.toml (image_dir = .../03_captioned/main). Nothing was added from unused candidates, and captions were not rewritten for this release.

This is a fan-made derivative, not an official product. Rights to the character and source work remain with their respective owners.

Contents

Item	Count
Training images (images/*.png)	80
Matching captions (images/*.txt)	80
Missing same-name captions	0
Orphan captions	0
num_repeats	2

Filenames are unchanged (for example 29.png / 29.txt). Captions are comma-separated Danbooru-style tags with the identity trigger fixed first:

emilia_rezero, 1girl, looking at viewer, smile, closed mouth, anime screencap

• Identity trigger: emilia_rezero only. There is no outfit token.
• Training used shuffle_caption = true and keep_tokens = 1, so only the first token stays pinned.
• Tags started from WD14, then every pair was manually reviewed. Identity words such as silver hair and purple eyes were left off the caption on purpose; the trigger is meant to carry them.
• The default look in this set is the white-and-purple ceremonial dress. Other clothes (travel cloak, pink cape dress, night gown) appear as ordinary prompt tags, not as extra trigger words.

Preview

These nine images are copies of real training stills, not model outputs. They cover face, three-quarter, profile, ceremonial dress, an alternate full-body outfit, sitting, travel coat, night, and forest.

Face	Three-quarter	Profile

Ceremonial dress	Alternate full body	Sitting

Travel coat	Night	Forest

How it was trained

• Base: waiIllustriousSDXL v17, 1024 buckets, bucket_no_upscale
• Network: LoRA rank 32 / alpha 16, UNet plus both text encoders
• Optimizer: AdamW8bit; UNet LR 1e-4, text-encoder LR 5e-5; cosine, 7 epochs, 1120 steps
• Recommended model from that run: E5 (Emilia_Rezero_Q-000005.safetensors) at strength 0.8

Related repositories

• Model: abbuibuibui/emilia-Lora — E1–E7 weights, bilingual card, and the public epoch × strength evaluation.

License

• Character and source-work rights remain with © Tappei Nagatsuki / White Fox / the relevant production committees.
• Dataset files are released under CreativeML OpenRAIL-M.
• Intended for non-commercial fan LoRA training and research.
• Do not use this dataset for unlawful, deceptive, harassing, or exploitative purposes.