Anime-faithful character LoRA for Senko-san from The Helpful Fox Senko-san, for use with FLUX.3 Image Klein Base (5.6B).Built from an entirely new dataset of 1,000 native 1024 × 1024 NovelAI 5.0 Curated generations, this ORIGINAL-ANIME version captures the look and feeling of the original TV anime. Each training image was generated to resemble a frame from an episode, emphasizing softer linework, familiar colors, and faithful character design for greater immersion.<h2 id="original-anime">ORIGINAL-ANIME?</h2>Where the HD version emphasizes ultra-crisp detail and high-fidelity rendering, ORIGINAL-ANIME prioritizes Senko’s on-screen likeness and the softer, screencap-like finish of the series. It is designed for <code>txt2img</code> character generation and <code>img2img</code> editing, with an emphasis on consistent identity and flexible natural-language prompting.This LoRA is an ULTRA variant built using the same prompt set and dataset structure as the previously released ULTRA variants, with all images newly generated in NovelAI 5.0 Curated to match the original TV anime’s look.Its 598-image core character set establishes Senko’s identity across varied angles, expressions, poses, and framing, while its 402 activity scenes expand that foundation with the camera angles, shot types, and poses found across a wide range of activities.<hr /><h2 id="recommended-starting-settings">Recommended Starting Settings</h2>Setting Recommendation Base model FLUX.3 Image Klein Base — 5.6B, undistilled Trigger word <code>senko_san</code> LoRA strength Start within <code>0.7–1.0</code> and adjust for likeness and flexibility CFG / guidance scale <code>3.5</code> Steps <code>50</code> Prompt upsampling Use BFL’s prompt upsampler or another LLM Negative prompt Leave completely emptyThe step count and guidance above follow the FLUX.3 Klein Base defaults. Treat them as a starting point when tuning this LoRA.<h2 id="prompting-guidance">Prompting Guidance</h2>Use a prompt upsampler for both generation and editing. FLUX.3 Klein expects a complete scene caption covering the subject, framing, environment, lighting, colors, and visual style.BFL’s <code>flux-3-image-klein-pup</code> expands your request into that caption. Another LLM can perform the same role. Make sure the resulting caption preserves the exact trigger word <code>**senko_san**</code>, along with the character details and composition you want.For <code>img2img</code>, provide the reference image to the upsampler together with your editing request. It should describe the complete intended final image, incorporating the changes while retaining the details you want preserved. A reference image helps anchor the content, but does not guarantee that the original composition will remain fixed.To reinforce the ORIGINAL-ANIME finish, describe its visible qualities: soft linework, familiar warm colors, simple cel shading, restrained highlights, and a TV-anime screencap appearance.<h3 id="character-colors">Character Colors</h3>Use these approximate color references when you want to reinforce Senko-san’s warm golden-yellow hair and fur:Feature Color reference Hair Warm golden yellow, approximately <code>#F2B93B–#F6C84A</code> Hair highlights Lighter gold, approximately <code>#FFD86A</code> Fox fur Warm golden yellow, approximately <code>#EFB53A–#F5C246</code> Fur accents and tail tip Clean whitePair the hexadecimal references with plain-language descriptions such as “rich, warm golden yellow” to communicate the intended palette.<h2 id="img2img-example">Img2img Example</h2>Provide the original image and the following request to your prompt upsampler. Pass its completed scene caption to FLUX.3 Klein with the reference image.<pre><code>Replace the character in the reference image with senko_san, using her recognizable face, eyes, short hairstyle, fox ears, and large fluffy fox tail.

Her short hair is rich, warm golden yellow, approximately #F2B93B to #F6C84A, with lighter golden highlights around #FFD86A. Her fox ears and tail have warm golden-yellow fur, approximately #EFB53A to #F5C246, with clean white fur accents and a white tail tip.

Use senko_san's correct short hairstyle in place of the original character's long hair. Her shoulders, arms, and hands have normal bare skin, including the area occupied by the original character's dark shoulder fur or shoulder patch.

Preserve the white dress, pose, camera angle, framing, composition, background, lighting, and environment shown in the reference. Preserve the large straw sun hat and sunflower decoration, with senko_san's fox ears visibly protruding through openings in the hat. Include her large fluffy tail in a position consistent with her pose and the scene.

Render the final image with soft linework, warm familiar colors, simple cel shading, restrained highlights, and the gentle screencap-like finish of the original TV anime.

Write a complete caption of the intended final image, including the preserved scene details visible in the reference. Retain the exact trigger word senko_san.
</code></pre>Recommended starting settings:<pre><code>CFG / guidance scale: 3.5
Steps: 50
LoRA strength: 0.7–1.0
Prompt upsampling: enabled, with the reference image
Negative prompt: empty
</code></pre><h2 id="txt2img-example">Txt2img Example</h2>Use the following scene description as input to your prompt upsampler, retaining <code>**senko_san**</code> in the final caption.<pre><code>senko_san appears in a full-color anime scene viewed from a first-person perspective, with a low camera angle and a slight Dutch tilt.

She has recognizable Senko-san facial features, short warm golden-yellow hair in the #F2B93B to #F6C84A range, lighter golden highlights around #FFD86A, and a red hair ornament. Her fox ears and large fluffy tail have warm golden-yellow fur in the #EFB53A to #F5C246 range, with clean white fur accents and a white tail tip.

She wears a cream summer sundress with a vermilion ribbon sash. She runs along a seaside carnival boardwalk, turning back over her shoulder while holding the viewer's hand and pulling them forward. She has a gentle affectionate expression, a small closed-mouth smile, and relaxed eyes. Her short hair, dress, and fluffy tail move in the sea breeze. The viewer's hand is visible in the foreground, warmly clasped in hers.

A glowing Ferris wheel rises behind her among the seaside amusement park attractions. Pink and magenta clouds stretch across a vivid sunset sky, with utility poles and a distant coastal city skyline completing the background. Warm golden sunset light falls across her face, accompanied by a soft magenta rim light.

The image has soft linework, simple cel shading, restrained highlights, warm familiar colors, and the gentle screencap-like appearance of the original TV anime. The composition feels lively, welcoming, and intimate, with Senko-san as the clear focal point.
</code></pre>Recommended starting settings:<pre><code>CFG / guidance scale: 3.5
Steps: 50
LoRA strength: 0.7–1.0
Prompt upsampling: enabled
Negative prompt: empty
</code></pre>

Senko-san_NAI-V5-Curated_ORIGINAL-ANIME_ULTRA (仙狐さん)

NovelAI 5.0 Curated Dataset - 1,000 ORIGINAL-ANIME style Images

Files with SHA256: fcf716267f7d12a3dcb291e552782d6b3d419209048a33c4e0e4685226a4a74e