Toji Fushiguro

Dipublikasikan pada a day ago

T S Tarun Rajput

character: fushiguro_toji, A full body shot from a low angle featuring Toji Fushiguro, a tall and imposing adult male standing confidently in a neutral, open environment. He possesses an extremely muscular athletic physique with broad shoulders, powerful arms, and a defined chest and narrow waist. His skin is lightly tanned, and he has a sharp, masculine face with a defined jawline, narrow intense eyes, and a distinctive scar extending across the right side of his mouth. His black hair is short, messy, and slightly spiky. He is dressed in a dark, tight-fitting shirt that accentuates his musculature, paired with loose dark pants and black shoes. His expression is calm yet intimidating, with a stoic, emotionless gaze that projects a dangerous presence. The scene is rendered as a high-quality anime illustration with clean linework and detailed anatomy. Dramatic lighting creates sharp focus and deep shadows, enhancing the three-dimensional feel of his physique against a minimalist background.

Tampilkan Parameter

Prompt Asli: 🐺 1. Character trigger words Use a unique trigger that is unlikely to occur naturally: toji_fushiguro_model For every training caption, include it near the beginning: toji_fushiguro_model, 1boy, adult male, ... If PixAI's training interface recommends a different trigger format, follow that format. 🔥 2. Core character prompt Use this as your base generation prompt after training: toji_fushiguro_model, Toji Fushiguro, 1boy, adult male, tall muscular man, extremely muscular athletic physique, broad shoulders, powerful arms, defined chest, narrow waist, strong neck, sharp masculine facial features, defined jawline, narrow intense eyes, black messy short hair, slightly spiky black hair, lightly tanned skin, distinctive scar extending across the right side of the mouth, calm intimidating expression, emotionless confident gaze, stoic demeanor, dangerous presence, physically imposing, dark fitted shirt, loose dark pants, black shoes, anime character, detailed face, detailed anatomy, clean linework, high quality, dramatic lighting, sharp focus ⚔️ For the classic Toji look toji_fushiguro_model, Toji Fushiguro, black fitted shirt, loose dark pants, muscular athletic body, broad shoulders, messy black hair, facial scar, calm intimidating expression, standing confidently, full body 🚫 3. Negative prompt I'd use a fairly aggressive negative prompt to protect the character's anatomy and identity: low quality, worst quality, low resolution, blurry, bad anatomy, bad proportions, malformed body, deformed hands, extra fingers, missing fingers, extra arms, extra legs, duplicate body, long neck, distorted face, asymmetrical eyes, child, teenager, young-looking, feminine, thin body, skinny body, weak physique, overly bulky body, exaggerated muscles, long hair, colored hair, beard, mustache, missing facial scar, incorrect facial scar, smiling, goofy expression, cute expression, chibi, cartoonish proportions, different character, duplicate character, text, watermark, logo, signature Important: don't overdo the negative prompt. If PixAI starts producing overly rigid results, remove some of the anatomy negatives. 📸 4. Training images — this matters MOST For a character LoRA/model, don't make every training image a dramatic fighting shot. I'd aim for roughly 15–25 high-quality images if your training setup allows it. Recommended dataset Image typeAmountPurposeFront face2–3Face identity3/4 face3–4Facial structureSide profile2Profile consistencyFull body4–5Body proportionsUpper body2–3PhysiqueDifferent poses3–4Pose generalizationDifferent expressions2–3Expression controlAction/combat2–3Dynamic appearanceDifferent camera distances2–3Composition flexibility Don't use 20 almost-identical screenshots. The model needs to learn what makes Toji Toji, rather than memorizing one pose. 📝 5. Recommended captions Keep captions descriptive but don't describe everything identically. Image 1 — front portrait toji_fushiguro_model, 1boy, adult male, muscular man, black messy short hair, narrow eyes, sharp jawline, facial scar across right side of mouth, serious expression, looking at viewer, portrait Image 2 — 3/4 portrait toji_fushiguro_model, 1boy, adult male, muscular athletic man, messy black hair, sharp masculine face, narrow intense eyes, facial scar, calm intimidating expression, 3/4 view, upper body Image 3 — full body toji_fushiguro_model, 1boy, adult male, tall muscular athletic man, broad shoulders, powerful arms, narrow waist, black messy hair, facial scar, dark fitted shirt, loose dark pants, standing, full body Image 4 — side profile toji_fushiguro_model, 1boy, adult male, muscular athletic physique, black messy short hair, sharp jawline, facial scar, side profile, serious expression Image 5 — combat toji_fushiguro_model, 1boy, adult male, extremely muscular athletic man, messy black hair, facial scar, intense focused expression, dynamic combat pose, powerful stance, dark clothing, full body Image 6 — relaxed toji_fushiguro_model, 1boy, adult male, muscular athletic man, messy black hair, facial scar, calm expression, relaxed posture, dark shirt, loose pants, standing 🧠 6. The important rule Don't put every characteristic into every caption. For example, if every single training image says: muscular, black shirt, black pants, standing the model may learn: Toji = black shirt + black pants + standing instead of: Toji = facial structure + scar + hair + physique + identity So vary: pose → camera → clothing → expression → environment while keeping the identity-defining features consistent. 🐺 7. Identity anchors These should remain extremely consistent throughout the dataset: HIGH PRIORITY Black messy short hair Sharp masculine face Narrow eyes Strong jaw Mouth scar Muscular athletic physique Broad shoulders Mature adult appearance VARIABLE Pose Camera angle Expression Clothing Background Lighting Action That gives you a model that can eventually generate: Toji sitting → Toji standing → Toji fighting → Toji walking → Toji portrait → Toji cinematic shot without losing his identity.
Mode: lite
Ukuran: 768 x 1280

Asisten prompt