Senko-san_NAI-V5-Curated_HD - Tsubaki.3

High-fidelity character LoRA for Senko-san from The Helpful Fox Senko-san, trained on Tsubaki.3.

Built from 100 native 1024 × 1024 NovelAI 5.0 Curated generations, this LoRA emphasizes sharp image quality, defined linework, solid anatomy, and very low compression blur. It captures Senko’s recognizable character design with crisp contours, clearly resolved details, and a polished illustration finish.

Tuned specifically for modern illustration fidelity, it supports txt2img, img2img, inpainting, and character head replacements. It is designed for both full character illustrations and targeted edits, with an emphasis on maintaining Senko’s likeness across different scenes and compositions.

Dataset

The training set combines 50 images from the original base dataset used to create my NovelAI 4.5 Curated LoRAs with 50 new activity scenes, balancing a strong character foundation with a broader range of situations.

The core character set anchors Senko’s identity, while the activity scenes expand the range of poses, camera angles, framing, and interactions represented in training. Together, they provide more versatility for everyday scenes and character-focused illustrations without sacrificing likeness or the sharp, detailed rendering that defines this release.