<h1 id="pink-axolotl-girl-or-character-lora">Pink Axolotl Girl | Character LoRA</h1>A cute anime-style Pink Axolotl Girl original character, available for both SDXL (via Animagine XL V3.1) and Anima (circlestone-labs/Anima-Base-v1.0). This LoRA is designed to reproduce the character's signature appearance while allowing flexibility in poses, expressions, lighting, and backgrounds.The character features pink hair, pink axolotl gills, a soft pastel aesthetic, and a scientist/laboratory-inspired outfit. Trigger Word (same for both versions) — <code>pinkaxolotlgirl</code><hr />SDXL v1.0 Base model: Animagine XL V3.1Recommended Prompt: <code>pinkaxolotlgirl, 1girl, pink hair, axolotl gills, pink eyes, white lab coat, black tank top, gray pants, masterpiece, best quality</code>Anima v1.0 Base model: Anima-Base-v1.0 — trained natively on the Anima architecture (not ported from SDXL), 768×768 dataset pass, taking advantage of Anima's language-model-based text encoder.Recommended Prompt: <code>pinkaxolotlgirl, 1girl, solo, pink hair, axolotl gills, pink eyes, white lab coat, black camisole, grey pants, looking at viewer, masterpiece, best quality</code> Recommended settings: LoRA weight 0.8–1.0, euler_a sampler, ~28 steps, CFG ~7.5, 768×768 generation resolution. Testing shows good identity consistency outside the training outfit/setting too — streetwear, winter clothing, and outdoor scenes all hold up well while keeping the character's core features intact. This is an early v0.1 release for Anima; a 1024×1024-trained version is planned to test whether higher training resolution improves fine detail without hurting flexibility. Anima v0.2 Update (1024×1024) Retrained at 1024×1024 (up from 768×768) to improve fine detail rendering. Sharper linework overall — hair strands, feather-ear texture, and shading transitions are noticeably crisper, along with more precise face detail (freckles, eye shading).Note: some generations show slight outfit drift toward a plain shirt/collar instead of the trained black camisole — prompting outfit details explicitly (e.g. <code>black camisole, v-neck</code>) helps pull it back on-model.Recommended generation resolution: 1024×1024 (matches this version's training res). Same trigger word and settings as v1.0 otherwise.<hr />Feedback is always welcome! If you notice anything that could be improved — character consistency, anatomy, prompting, or overall quality — please leave a comment on either version. Every suggestion helps me improve future updates of this LoRA.

Pink Axolotl Girl Character LoRA

Pink Axolotl Girl — Anima LoRA (v0.1)A character LoRA for the Anima base model (circlestone-labs/Anima-Base-v1.0), trained on a custom 59-image dataset with danbooru-style tag captions.Trigger word: <code>pinkaxolotlgirl</code>Training details:<ul><li>Base model: Anima-Base-v1.0 (Qwen3-0.6B text encoder, Qwen Image VAE)</li><li>Resolution: 768×768</li><li>Network: LoRA, rank 32 / alpha 16</li><li>~3,540 steps, 6 epochs, danbooru tag captions with caption shuffle disabled (text encoder outputs cached)</li></ul>Recommended usage:<ul><li>LoRA weight: 0.8–1.0</li><li>Sampler: euler_a, ~28 steps, CFG ~7.5</li><li>Works well combined with outfit/setting/pose tags outside the training data (streetwear, winter, casual, dynamic poses all tested and hold identity well)</li></ul>Notes: This is an early (v0.1) test release. A second version trained at 1024×1024 is planned to evaluate whether higher training resolution improves fine detail (face, fins) without hurting flexibility.

Files with SHA256: 7d8b7689e008aa23f1ee90d21ebb4036a90022f6c658a24c26ade15fee7c2650