eight tohyama style LoRA for Anima (base v1.0) and Anima-2.9BA Danbooru-tag LoRA trained to replicate the art style of the illustrator @eight tohyama — clean linework, soft cel shading, delicate facial features, and light pastel palettes with detailed hair rendering.Works on both the original Anima base v1.0 and the new Anima-2.9B (40-layer expanded) model.### Trigger words```@eight tohyama```Put it at the very front of your prompt (keep-token 0). The tag also blends well with other artist tags for style mixing.### Supported charactersTrained on the artist's original character roster — works with the following character names (Danbooru tags, use as-is):| Character | Tags to add ||---|---|| anri (main, 10 imgs) | <code>anri</code> — aqua blue eyes, small breasts, blue hair clip, long hair || saito | <code>saito</code> — black hair side ponytail, pink eyes || suou | <code>suou</code> — black hair, short twintails, blunt bangs, brown eyes, loli/petite || Graf Zeppelin (Kancolle) | <code>graf zeppelin (kancolle)</code> — blonde, iron cross, kancolle uniform |Character name goes right after the artist tag. For copyright characters, follow the name with the series tag <code>kancolle</code>).### Versions| Version | Base model | File | Size ||---|---|---|---|| 1.0 (base) | <code>anima-base-v1.0.safetensors</code> | <code>eight_tohyama_base_1.0.safetensors</code> | 132 MB || 1.0 (2.9B) | <code>Anima-2.9B-preview-v1.safetensors</code> | <code>eight_tohyama_1.0.safetensors</code> | 172 MB |&gt; ⚠️ Version-specific: the 2.9B build is trained on the 40-layer expanded architecture and is not compatible with base v1.0 (and vice versa). Use the matching file for your base model.### Training details- Dataset: 33 images of eight tohyama artwork (Danbooru tags, 1024² resolution, 4 repeats)- Trained with kohya sd-scripts <code>anima_train_network.py</code>)- LoRA dim 16 / alpha 16, LR 3e-5, 8-bit AdamW, bf16, constant schedule, seed 42- base 1.0 build: 40 epochs → best at epoch 30- 2.9B build: 30 epochs → best at epoch 22- Bucket training enabled; no regularisation images### Recommended settingsAnima base v1.0 (with this LoRA at strength 1.0):| Parameter | Value ||---|---|| Sampler | <code>er_sde</code> (or <code>euler_a</code>) || Scheduler | <code>simple</code> (or <code>beta57</code>) || Steps | 25–30 || CFG | 4.0–4.5 || Flow shift | 3.0 || Resolution | 888×1184 / 1024² / up to 1152×1536 || LoRA strength | 0.8–1.0 (1.0 default; lower for subtler style) |Anima-2.9B (with this LoRA at strength 1.0):| Parameter | Value ||---|---|| Sampler | <code>euler</code> / <code>res-multistep</code> / <code>er-sde</code> || Scheduler | <code>sgm-uniform</code> / <code>beta</code> / <code>linear-quadratic</code> || Steps | 28–50 (sweet spot ~30) || CFG | 3.5–5 || Flow shift | 2.0–3.0 || Resolution | 812×1216 / 1152×1536 / 16:9 up to 1496×840 || LoRA strength | 0.8–1.0 |### Tips- Detailed prompts work best; the style tag blends with character tags, series tags and other artist tags- For face-focused portraits, 3/4 view reduces top whitespace; avoid <code>frame</code> / <code>out of frame</code> wording (triggers picture-in-picture artifacts on 16:9)- Negative prompt starters: <code>worst quality, low quality, artist name, blurry, watermark, text</code>- No score tags needed; the model responds to quality tags <code>masterpiece, best quality</code>)### Notes- Derivative of [circlestone-labs/Anima](<a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima">https://huggingface.co/circlestone-labs/Anima</a>) — non-commercial license applies to generated content and redistribution. Mark this LoRA as Non-Commercial on Civitai.- Training data captions are Danbooru-style tags; keep-token 0 keeps the artist tag at the prompt head.

Files with SHA256: c9a145ff4eefd28e28b0154a4a0a4a80cd28bf8033d1943f0ae48d633f095da3