Model
Trends
.ai
Model
Trends
.ai
les meilleurs modèles d'IA open source

Krea2-Anime v1.5 : un fine-tune anime qui récompense les prompts détaillés

Sujet : Modèles
Par Captain
Publié le 2026-10-02
Rendus d'exemple
Prompts et réglages partagés par les personnes qui ont réalisé ces rendus sur Civitai. Choisissez un rendu pour voir comment il a été fait.
Réglages
Sampler
Euler
Étapes
8
Guidance
1
Seed
1440015573
Taille
1536x2688
Prompt
masterpiece, very aesthetic, best quality, score_7, score_9, sensitive, ultra detailed, --preview, in ANIRL Style, Studio Ghibli Dark Fairytale, dynamic pose, depth of field, anime illustration of a young woman with short black bob hair and blunt bangs, wearing black cat-ear shaped hair clips, sharp brown eyes looking directly at viewer, leaning her shoulder and back casually against a weathered brick wall on one side of a narrow alleyway, off-center asymmetrical composition, the opposite side of the alley open and receding into hazy blurred cityscape with red signage in the distance. She wears an elaborate oversized punk/techwear hooded jacket in black and red, with fur trim around the hood, multiple zippers, metal buckles, chain details. Across the sleeves and chest, a large faded red fabric panel with the bold white lettering "SIA" appliqued on top, the white letters edged with a visible zigzag overlock stitch along their borders, the red fabric muted and slightly desaturated rather than bright red, showing worn texture with frayed edges, small scuffs, cracked and peeling patches, and uneven fading as if weathered from long use. Underneath, a red plaid pleated skirt and black fingerless gloves with buckle straps, dark thigh-high socks. Highly dynamic twisted pose, one hip popped out, torso turned toward viewer while shoulder rests against the wall, one arm bent up near her face gripping her jacket collar, the other arm hanging loose with fingers grazing the wall, weight shifted onto one leg, fabric and chains caught mid-motion as if she just leaned back, wind-blown strands of hair, confident and edgy expression. Extremely strong rim lighting tracing the entire silhouette of her hair, shoulders, and jacket fur in blazing warm gold, intense backlight flare blooming directly behind her from low in the frame, near-silhouette shadow on the front of her body with only key facial features and eyes catching a touch of light, thick atmospheric haze scattering the light, strong bloom and lens flare artifacts, cinematic depth of field with blurred foreground elements at the alley edges. Anime illustration style, detailed line art, painterly cel-shading, extreme high contrast between blazing backlight and deep shadow, vertical portrait composition, masterpiece quality, intricate fabric and clothing details, worn distressed texture rendering.
Réglages
Sampler
Euler
Étapes
8
Guidance
1
Seed
3987435268
Taille
1536x2688
Prompt
masterpiece, very aesthetic, best quality, score_7, score_9, sensitive, ultra detailed, --preview, in ANIRL Style, zidiusArt, dynamic pose, depth of field, anime illustration of a young woman with short black bob hair and blunt bangs, wearing black cat-ear shaped hair clips, sharp brown eyes looking directly at viewer, leaning her shoulder and back casually against a weathered brick wall on one side of a narrow alleyway, off-center asymmetrical composition, the opposite side of the alley open and receding into hazy blurred cityscape with red signage in the distance. She wears an elaborate oversized punk/techwear hooded jacket in black and red, with fur trim around the hood, multiple zippers, metal buckles, chain details. Across the sleeves and chest, a large faded red fabric panel with the bold white lettering "SIA" appliqued on top, the white letters edged with a visible zigzag overlock stitch along their borders, the red fabric muted and slightly desaturated rather than bright red, showing worn texture with frayed edges, small scuffs, cracked and peeling patches, and uneven fading as if weathered from long use. Underneath, a red plaid pleated skirt and black fingerless gloves with buckle straps, dark thigh-high socks. Highly dynamic twisted pose, one hip popped out, torso turned toward viewer while shoulder rests against the wall, one arm bent up near her face gripping her jacket collar, the other arm hanging loose with fingers grazing the wall, weight shifted onto one leg, fabric and chains caught mid-motion as if she just leaned back, wind-blown strands of hair, confident and edgy expression. Extremely strong rim lighting tracing the entire silhouette of her hair, shoulders, and jacket fur in blazing warm gold, intense backlight flare blooming directly behind her from low in the frame, near-silhouette shadow on the front of her body with only key facial features and eyes catching a touch of light, thick atmospheric haze scattering the light, strong bloom and lens flare artifacts, cinematic depth of field with blurred foreground elements at the alley edges. Anime illustration style, detailed line art, painterly cel-shading, extreme high contrast between blazing backlight and deep shadow, vertical portrait composition, masterpiece quality, intricate fabric and clothing details, worn distressed texture rendering.
Réglages
Sampler
Euler
Étapes
8
Guidance
1
Seed
3506206289
Taille
1664x2432
Prompt
masterpiece, very aesthetic, best quality, score_7, score_9, sensitive, ultra detailed, --preview, zidiusArt, PoseDaal, depth of field, Extreme fisheye wide-angle lens shot with strong spherical barrel distortion warping the entire frame, the character positioned close to the camera and facing directly toward the viewer, as if the viewer is looking straight up at her from below during her fall. Anime key visual illustration of a young woman falling through the sky from extremely high altitude — this is a high-altitude atmospheric view looking down at cloud layers far below, NOT outer space, no stars, no galaxy, no cosmic void, no distant planet. The scene shows a vast sea of clouds beneath her, a hazy pale-blue sky fading toward a bright horizon line, with the sun blazing low near that horizon, creating a slight backlit/rim-light composition where sunlight glows around the edges of her hair and figure while her front is softly shadowed. The character wears a classic maid outfit: a black top with white apron and puffed white long sleeves, and a white headpiece/hairband, with short bob-cut hair and aviator goggles pushed up onto her forehead. Her maid skirt is unusually long and voluminous, a full flowing floor-length skirt with layered ruffled fabric, and because of the extreme high-altitude wind, this long skirt is violently whipping and billowing outward, twisting and rippling dramatically in the air around her legs. She grips a long straight sword with both hands. Two ribbons, one pink and one white, are tied to the end of the handle, trailing and streaming far behind her in the wind alongside her hair and skirt. Her body is in a backswing motion, torso and shoulders twisted and coiled as if winding up to swing the sword forward with full force in the next instant. Her expression is a confident, focused smirk, eyes looking directly at the viewer, hair, apron, and the long billowing skirt all whipping upward and outward from the speed of the fall and high-altitude wind. Bold clean anime line art, crisp cel shading with sharp highlight and shadow separation, vivid saturated color palette dominated by sky blue, white, and black with warm golden-orange backlight glow from the low sun, dramatic rim lighting, sharp focus, high contrast, dynamic action illustration, ultra-detailed rendering.no rear view,
Réglages
Sampler
Euler
Étapes
8
Guidance
1
Seed
579296002
Taille
1024x768
Prompt
anime illustration style two adult female cheerleaders standing together on the campus of a prestigious American college, two clearly distinct women, different facial features, different hairstyles, different makeup, different clothing colors, individually recognizable identities, full-body composition, both women clearly visible from head to feet, standing close enough to appear together while maintaining separate identities. LEFT FEMALE: rlygenoveva, platinum blonde hair, brown eyes, realistic eyes, adult female, about 25 years of age. Her long platinum blonde hair is styled in a high athletic ponytail, with the hair falling in smooth thick sections behind her shoulders. Several loose strands frame her face naturally. She has realistic facial proportions, softly defined cheekbones, a feminine jawline, naturally shaped full lips and an elegant nose. Her skin shows natural texture, visible pores and subtle microdetail. Her brown eyes, realistic eyes, have detailed iris patterns with warm chestnut tones and subtle golden variations around the pupils. Natural catchlights reflect the daylight, while the whites of her eyes remain clean and clearly defined. Her makeup is fresh and polished with softly blended warm-toned eyeshadow, defined eyelashes, lightly shaped eyebrows, subtle blush and glossy rose-colored lips. Her expression is energetic, confident and friendly. She looks directly toward the camera with a bright natural smile, slightly raised cheeks and expressive eyes. Her posture communicates enthusiasm and team spirit without appearing exaggerated, natural_expressions. She wears a modern college cheerleading uniform in deep navy blue, white and metallic silver. The fitted cropped athletic top has clean geometric panels and a small college-style emblem, combined with a matching pleated mini skirt with layered fabric panels. White athletic sneakers complete her outfit. She holds metallic silver and navy pom-poms naturally at her sides. Her fingernails are medium-length, professionally manicured and painted glossy navy blue with thin silver accent lines. She stands on the left side of the composition with her body angled slightly toward the center. Her weight rests on one leg while the other leg is positioned naturally forward. One arm is relaxed beside her body holding a pom-pom, while her other hand rests confidently against her waist. RIGHT FEMALE: rlyfiona, honey brown hair, light freckles, adult female, about 24 years of age. Her honey brown hair is styled completely differently, arranged in two high braided ponytails with loose wavy sections flowing from the braids. The hairstyle creates a youthful and energetic appearance while remaining suitable for an athletic college cheerleader. Light freckles are naturally visible across her nose and upper cheeks. She has distinct facial features from the woman on the left, with a slightly different face shape, naturally defined cheekbones, a softer jawline and expressive lips. Her skin has realistic texture, visible pores and subtle microdetail. Her gray-blue eyes, realistic eyes, display detailed iris structure with cool blue and soft gray variations surrounding the pupils. A subtle darker limbal ring defines the edge of each iris, with natural catchlights reflecting the outdoor campus environment and clean whites of the eyes. Her makeup is noticeably different from the woman on the left. She wears a more playful contemporary look with softly defined eyeliner, light peach eyeshadow, naturally shaped eyebrows, subtle highlighter across the cheekbones and glossy coral-colored lips. Her freckles remain clearly visible beneath the makeup. Her expression is playful and slightly mischievous. She has a restrained asymmetric smile, with one corner of her lips raised slightly higher than the other. Her eyes look toward the camera with a lively, confident spark. One eyebrow is subtly raised, creating a more spontaneous and individual expression. Her facial muscles remain relaxed and believable, natural_expressions. She wears a different modern cheerleading uniform in burgundy red, cream white and subtle gold details. Her fitted athletic top has a different cut from the first woman, with crossed straps and contrasting sleeve details. She wears a matching pleated cheerleading skirt with burgundy and cream panels. White and burgundy athletic sneakers complete the outfit. She holds gold and cream pom-poms, visually separating her team colors from the woman on the left. Her hands are clearly visible and naturally positioned. Her fingernails are medium-long and professionally manicured, painted glossy burgundy red with small cream-colored accent details on several nails. She wears small gold hoop earrings and a thin delicate bracelet. She stands on the right side of the composition with a noticeably different body posture. Her body is angled slightly outward while her upper body turns back toward the camera. One knee is relaxed, creating a more dynamic stance. One arm is slightly raised with a pom-pom resting near her shoulder, while her other arm hangs naturally beside her body holding the second pom-pom. BACKGROUND: a prestigious American college campus with large historic university buildings, red brick architecture, broad green lawns, mature trees and a recognizable college sports field visible in the distance. Students walk across the campus behind them, while university banners and campus architecture establish the American college atmosphere. The environment remains secondary to the two women. The two women occupy the foreground and remain the dominant subjects of the composition. Left female and right female remain clearly separated and individually recognizable, with different faces, hairstyles, hair colors, eye colors, makeup, uniforms, manicure colors, jewelry and body language. Camera positioned close enough to capture both women in detailed full-body portraits, central composition, both subjects prominent in the frame, detailed faces, realistic eyes, detailed hands, complete uniforms and individual body postures clearly visible. 35mm photography, contemporary college lifestyle photography, bright natural daylight, realistic depth of field, detailed campus background, natural skin texture, visible pores, skin microdetail, soft daylight, soft bounce light, subtle rim light, realistic shadows, expressive eyes, detailed hands, realistic athletic fabric textures, natural_expressions. <lora:training_3876538-20260808162747105:0.5> <lora:detail_slider_krea2_loraholic:2> <lora:wetness_krea2_loraholic:-1.3> <lora:krea2_More_Expressive_And_Best:0.8> <lora:training_2380271-20260827204401576:0.8> <lora:ThotShot-ZiB-genoveva-trigger_copy:1> <lora:thotshot-K2-sarahhd-trigger:1> , clean anime lineart, smooth confident outlines, soft cel shading, balanced anime proportions, expressive anime eyes, clear iris structure, subtle highlights, simplified nose and lips, painterly anime shading, flat yet rich color palette, gentle gradients, controlled lighting, studio illustration look, high visual clarity, centered or rule-of-thirds composition, stable anatomy, clean silhouette, modern high-quality anime artwork
Prompt négatif
photorealistic, photo, realism, hyperrealism, ultra detailed skin, pores, wrinkles, skin texture, noisy, grainy, messy lines, sketchy, painterly oil, 3d render, CGI, doll face, plastic skin, overrendered, excessive shading, harsh shadows, HDR, bad anatomy, deformed face, extra limbs, mutated hands
Krea2-Anime de youneedanewname transforme Krea 2 en modèle anime, mais il brille surtout avec des descriptions de scènes riches. Découvre sa philosophie de prompt, la structure tag plus prose, les réglages et des idées.

Aperçu

Krea2-Anime (v1.5) est un fine-tune
anime
de
Krea 2
encore en cours d'entraînement (environ un quart terminé à cette version) mais déjà fonctionnel. Le point central de l'auteur : les prompts sont essentiels. Les prompts simples donnent un anime générique ; des descriptions de scènes très détaillées donnent un résultat d’un autre niveau.
Approche recommandée : un mélange de langage naturel riche (environnement, action, personnage, style) avec des tags stratégiques. Le tag ANIRL Style du catalogue apparaît dans ses exemples.

Référence

Nom
Type
Rôle
Prompt
prose + tags
Descriptions longues et détaillées de la scène.
Steps / guidance
8 / 1
Réglages par défaut de Krea 2.
Style tags
in ANIRL Style, dynamic pose, depth of field
Quality tags
masterpiece, very aesthetic, best quality
Optionnel.
LoRAs
Krea 2 LoRAs
Les sliders fonctionnent.
AD
Aucune installation requise – exécutez les modèles IA dans le cloud
BitVector Prism est la façon la plus simple de commencer : choisissez un modèle, tapez un prompt et générez dans une application web épurée, sans rien configurer. BitVector est aussi disponible sur Discord et sur le web (SpyGlass).

Pas à pas

  1. Écris un paragraphe complet : perspective, éclairage, détails du personnage, décor, style.
  2. Ajoute quelques tags (dynamic pose, depth of field).
  3. 8 steps, guidance 1.
  4. Si le résultat paraît générique, ajoute plus de détails sur la scène, pas plus de tags qualité.
  5. Utilise le slider Detail à +0.5 pour la texture.

Exemples

Dark fantasy

An anime-style digital illustration in a dark fantasy gothic aesthetic with a dramatic low-angle perspective emphasizing depth and scale, an adult demon woman with long flowing white hair and large curved black horns seated on an elevated stone ledge, heterochromia (one glowing red eye, one cyan), pointed ears, cathedral ruins behind, moonlight. masterpiece, very aesthetic, best quality

Style Ghibli

masterpiece, very aesthetic, best quality, in ANIRL Style, dark fairytale, dynamic pose, depth of field. An anime illustration of a young woman with a short black bob and a lantern walking through a glowing mushroom forest at night, mist, fireflies.

Idée créative : fenêtre de train

A detailed anime illustration from inside a night train, an adult traveler asleep against the window, reflections of passing city lights on the glass, warm cabin light, rain streaks, depth of field.

Astuces

  • La perspective en contre-plongée et les indices de profondeur sont respectés précisément.
  • Décris les couleurs et matériaux des yeux et des cheveux ; il les rend bien.
  • Les esthétiques gothique et dark fantasy sont un point fort.
  • Les références style studio (conte de fées sombre) influencent le rendu.

Dépannage

Résultat anime générique

Pourquoi cela arrive
Prompt trop court.

Comment le corriger
Ajoute des détails sur l’environnement, l’action, le personnage et le style.

Erreurs d’anatomie

Pourquoi cela arrive
Phase d’entraînement précoce.

Comment le corriger
Poses plus simples ; régénère.

AD
PirateDiffusion
Aucune installation requise – exécutez les modèles IA dans le cloud
PirateDiffusion est réservé à Telegram et conçu pour les pros : des milliers de modèles, LoRA et workflows pilotés par commandes de chat, avec génération illimitée à prix fixe.

Questions

Terminé ?

Non, l’entraînement continue ; v1.5 est un instantané.

LoRAs ?

Les
LoRAs
et sliders de Krea 2 fonctionnent.

Liens et sources

Modèles de ce guide

Écrit par
Captain

Guides associés

Modèles
Curseur de luminosité pour Krea 2 : exposition en nombre
Le curseur de luminosité de Loraholic modifie l'exposition de toute l'image dans Krea 2, plus sombre aux valeurs positives. Découvre la plage de -5 à 5, comment il diffère du curseur Light et quand il vaut mieux utiliser un prompt.
Par Captain
2026-09-29
Modèles
Anything2RealCharacters V3 pour Krea 2 : transforme l'art 2D en personnes photoréalistes
Anything2RealCharacters V3 de AIGC_Singularity est un LoRA de stylisation image-à-image qui convertit des entrées anime et artistiques en personnages photoréalistes tout en gardant l'identité. Découvre le processus, le débruitage et les idées.
Par Captain
2026-10-03
Modèles
Style Anime Plat pour Krea 2 : cel-shading 2D audacieux
Le LoRA Anime Flat Style d'Agino donne à Krea 2 des œuvres d'art anime nettes avec des couleurs plates audacieuses et un ombrage minimal. Apprends le poids, la structure des prompts et des idées pour les designs de personnages et les scènes style capture d'écran.
Par Captain
2026-10-02
Modèles
Anime vers Réalisme Krea2Edit : convertir des images d'anime en photos en prise de vue réelle
Anime to Realism de MengY_Madrigal est un LoRA Krea 2 Edit qui transforme des images d'anime en photos photoréalistes ou de cosplay haut de gamme tout en gardant la pose, la composition et le costume. Découvre la configuration des nœuds, le réglage KV cache et les deux prompts officiels.
Par Captain
2026-10-03

Autres guides

Modèles
FLUX.1 Dev : comment le guider, meilleurs réglages et idées créatives
Un guide pratique pour FLUX.1 Dev de Black Forest Labs : comment utiliser des instructions en langage naturel, réglages de guidage et d’étapes, empilement de LoRA, rendu de texte et idées de prompts qui exploitent ses points forts.
Par Captain
2026-09-30
Modèles
Stable Diffusion XL 1.0 : prompts, réglages et ce qu'il fait encore de mieux
Comment tirer le meilleur parti du modèle de base officiel SDXL 1.0 : résolutions natives, réglages CFG et sampler, le refiner, prompts négatifs et idées de styles où SDXL brille encore.
Par Captain
2026-10-01
Modèles
Stable Diffusion 1.5 : le modèle classique, bien utilisé
Stable Diffusion 1.5 vaut toujours le coup : la bonne résolution, CFG et sampler, comment utiliser sa grande bibliothèque de LoRAs et embeddings, et des idées de prompts créatifs adaptés à un modèle 512 pixels.
Par Captain
2026-10-02
Modèles
FLUX.2 Dev : comment utiliser le nouveau modèle de Black Forest Labs
FLUX.2 Dev permet des prompts plus longs, un meilleur texte, un réalisme renforcé et une édition multi-références. Ce guide explique le workflow turbo, les réglages de guidance et de résolution, la structure du prompt et des idées qui tirent parti de ses nouvelles capacités.
Par Captain
2026-10-03
Modèles
Qwen-Image : longues invites, texte parfait et affiches bilingues
Qwen-Image est le modèle à utiliser quand les mots dans l’image comptent. Apprends à écrire ses longues invites descriptives, à rendre le texte anglais et chinois avec précision, à régler le CFG et les étapes, et à explorer des idées créatives avec beaucoup de mise en page.
Par Captain
2026-09-29
Modèles
SDXL-Lightning : génération en quatre étapes sur n'importe quel checkpoint SDXL
Le LoRA SDXL-Lightning de ByteDance transforme un rendu SDXL de 30 étapes en un rendu de 4 étapes. Découvre le nombre d'étapes, le CFG à utiliser, le sampler, et comment le combiner avec tes checkpoints et LoRAs de style préférés.
Par Captain
2026-09-30