本文へ移動
ぽちめのLINEスタンプサイト

ナヒーダのAI画像生成プロンプト5選|ポーズ・背景・画角を変えた結果比較

約16分
ナヒーダのAI画像生成プロンプト5選|ポーズ・背景・画角を変えた結果比較 — 『原神』の人気キャラ候補からナヒーダをランダム選定し、森林への落下、砂漠サーフィン、逆さ図書館、月夜のダンス、遺跡でのしゃがみ構図を生成。似た絵を量産しないための英語・日本語プロンプトとネガティブプロンプトを公開します。

『原神』の人気キャラ候補からナヒーダをランダム選定し、森林への落下、砂漠サーフィン、逆さ図書館、月夜のダンス、遺跡でのしゃがみ構図を生成。似た絵を量産しないための英語・日本語プロンプトとネガティブプロンプトを公開します。

同じキャラクターを何度生成しても、ポーズや背景、光が似ていれば投稿一覧では1枚に見えてしまいます。今回はナヒーダの髪・瞳・衣装だけを共通条件として固定し、動作、場所、時間帯、カメラ、感情を5案すべて別設計にしました。実際の生成結果を使い、どの指定が画像の差へ効いたかを比較します。

この記事で分かること

  • 1ナヒーダを選んだランダム選定の考え方
  • 2ポーズと背景が重複しない5つの実用プロンプト
  • 3森林・砂漠・図書館・湖・遺跡の生成結果
  • 4ローアングル・横追い・魚眼俯瞰・正面遠景・近接構図の使い分け
  • 5年齢にふさわしい表現と破綻を抑えるネガティブプロンプト
  • 6SNSで目を止める画像を選ぶための比較基準

1似た画像を避けるために固定した要素・変えた要素

キャラクターの識別条件まで毎回変えると、5枚の比較ではなく別キャラの比較になります。そこで銀白色の髪、緑の毛先、尖った耳、葉形の緑の瞳、白緑と金の衣装、健全で年齢にふさわしい表現は固定しました。一方、ポーズ、背景、時間帯、画角、主な動線、表情は重複を禁止するレベルで書き分けています。

  • 森林:自由落下 × 極端なローアングル × 明るい朝
  • 砂漠:サンドサーフィン × 横追い × 天気雨
  • 図書館:逆さの無重力 × 俯瞰魚眼 × 真夜中
  • 湖:片足のダンス × 正面遠景 × 月光
  • 遺跡:巨大な手の上でしゃがむ × 近接ローアングル × 日の出

5案で共有できるネガティブプロンプト

各生成プロンプトのAvoid欄にも個別の除外条件を入れています。下記はキャラ崩れ、年齢表現、手足、トリミング、構図の重複をまとめて抑えたい場合の共通版です。

incorrect character, wrong hair color, missing pointed ears, adult proportions, sexualized pose, revealing clothing, cleavage, static standing pose, repeated background, repeated camera angle, duplicated character, extra limbs, extra fingers, fused hands, broken wrists, twisted joints, cropped head, cropped feet, hidden face, unreadable eyes, photorealism, 3D render, muddy colors, blur, low contrast, text, captions, logo, watermark, signature

3ナヒーダのポーズ・背景を変えた5つの生成結果

プロンプト欄は英語が初期表示です。日本語ボタンで翻訳へ切り替えられます。長いプロンプトは隣の画像と同じ高さの範囲でスクロールできるため、画像と指定内容を見比べてください。

結果1:森林上空の自由落下

サムネ映え
スメールの森林上空を自由落下し、光る葉へ手を伸ばすナヒーダのAI生成イラスト

森林・落下プロンプト

Use case: stylized-concept
Asset type: vertical social-media anime illustration and blog experiment result, first image in a five-scene series
Primary request: Create a premium Japanese 2D anime key visual of Nahida from Genshin Impact, faithfully recognizable with her small childlike proportions, pale silver-white bobbed hair with soft green-tinted ends, long pointed elf-like ears, bright green leaf-shaped eyes, and ornate white-and-green Dendro dress with gold and leaf motifs. Keep the character wholesome, nonsexualized, and age-appropriate.
Scene/backdrop: high above Sumeru's lush rainforest canopy at bright morning, enormous luminous tree branches below, waterfalls and mist far in the distance, scattered glowing Dendro leaves rushing upward.
Subject/action: Nahida is joyfully free-falling through the open canopy, body tilted headfirst on a strong diagonal, one hand reaching toward a luminous leaf below and the other arm swept behind her, legs bent naturally as ribbons and capelets stream upward. She looks toward the viewer with an adventurous, delighted expression.
Style/medium: polished commercial Japanese anime illustration, crisp controlled line art, refined cel shading, highly detailed but readable costume, cinematic depth, clean facial features.
Composition/framing: 4:5 portrait, full body visible, extreme low-angle wide-lens view looking up through the branches, dramatic depth and foreshortening, face unobstructed, asymmetrical diagonal silhouette, strong thumbnail readability.
Lighting/mood: fresh emerald daylight, white sun shafts, cool turquoise mist, exhilarating and airy.
Color palette: leaf green, pearl white, turquoise, pale gold, and sky blue.
Constraints: only one Nahida; accurate childlike facial and body proportions; natural hands and limbs; no text, no logo, no watermark, no signature.
Avoid: standing pose, sitting pose, desert, indoor library, moonlit lake, ruins, duplicate character, sexualized clothing or pose, adult body, cleavage, extra limbs, extra fingers, fused hands, cropped head, cropped feet, hidden face, photorealism, 3D render, muddy colors, blur, cluttered composition.

結果2:砂漠の葉サーフィン

速度感
天気雨のスメール砂漠で光る葉に乗り、砂丘を横滑りするナヒーダのAI生成イラスト

砂漠・横移動プロンプト

Use case: stylized-concept
Asset type: vertical social-media anime illustration and blog experiment result, second image in a five-scene series
Primary request: Create a premium Japanese 2D anime key visual of Nahida from Genshin Impact, faithfully recognizable with her small childlike proportions, pale silver-white bobbed hair with soft green-tinted ends, long pointed elf-like ears, bright green leaf-shaped eyes, and ornate white-and-green Dendro dress with gold and leaf motifs. Keep the character wholesome, nonsexualized, and age-appropriate.
Scene/backdrop: a vast Sumeru desert during a dramatic sun shower, sweeping golden dunes, distant sandstone ruins, dark teal storm clouds split by warm sunlight, translucent Dendro symbols scattering like green sparks across the sand.
Subject/action: Nahida is sand-surfing sideways down a steep dune on a broad glowing leaf made of Dendro energy. She crouches low like a skilled board rider, one arm extended backward for balance and the other hand skimming near the sand, sending a long curved spray of gold grains and green light behind her. Her expression is focused, playful, and fearless.
Style/medium: polished commercial Japanese anime illustration, crisp controlled line art, refined cel shading, detailed but readable costume, cinematic environmental depth.
Composition/framing: 4:5 portrait, full body visible in clear side profile, lateral tracking-camera feel, long S-curve sand trail across the frame, horizon strongly tilted, no foreground reaching hand, face visible in three-quarter profile, dynamic negative space ahead of her movement.
Lighting/mood: hot amber sunlight against cool blue-green storm light, adventurous, fast, expansive.
Color palette: desert gold, amber, teal-gray, leaf green, pearl white.
Constraints: only one Nahida; accurate childlike facial and body proportions; natural hands and limbs; clearly different from a forest falling scene; no text, no logo, no watermark, no signature.
Avoid: forest canopy, falling pose, indoor library, upside-down floating, moonlit lake, ruins as the main foreground, static standing, frontal pose, adult body, sexualization, cleavage, duplicate character, extra limbs, extra fingers, fused hands, cropped head, cropped feet, photorealism, 3D render, blur, muddy colors.

結果3:真夜中の逆さ図書館

構図の意外性
真夜中の円形図書館で逆さに浮かび、光る草元素紋様を描くナヒーダのAI生成イラスト

図書館・逆さ浮遊プロンプト

Use case: stylized-concept
Asset type: vertical social-media anime illustration and blog experiment result, third image in a five-scene series
Primary request: Create a premium Japanese 2D anime key visual of Nahida from Genshin Impact, faithfully recognizable with her small childlike proportions, pale silver-white bobbed hair with soft green-tinted ends, long pointed elf-like ears, bright green leaf-shaped eyes, and ornate white-and-green Dendro dress with gold and leaf motifs. Keep the character wholesome, nonsexualized, and age-appropriate.
Scene/backdrop: inside a vast circular Sumeru Akademiya library after midnight, concentric balconies, dark carved bookshelves, suspended lamps, hundreds of open books and loose pages orbiting through emerald dream light.
Subject/action: Nahida floats weightlessly upside down at the center of the library atrium, knees softly tucked, one hand holding a tiny glowing seed close to her face and the other tracing a circular Dendro glyph. Her hair, ribbons, pages, and dress panels drift in different directions, making zero gravity obvious. Her expression is calm, curious, and absorbed in discovery.
Style/medium: polished commercial Japanese anime illustration, crisp elegant line art, refined cel shading, ornate architectural detail, luminous magical effects, readable character silhouette.
Composition/framing: 4:5 portrait, complete body visible, dramatic top-down fisheye view into the circular atrium, Nahida visibly upside down, strong concentric spiral layout, no forward-reaching hand, face unobstructed, subject centered but surrounded by deep architectural rings.
Lighting/mood: midnight indigo shadows, emerald book-light, small warm amber lamps, mysterious, intelligent, dreamlike, quiet rather than action-heavy.
Color palette: deep indigo, emerald green, parchment cream, dark wood, restrained warm gold.
Constraints: only one Nahida; accurate childlike facial and body proportions; natural hands and limbs; clearly different from outdoor forest and desert action scenes; no text, no readable writing on pages, no logo, no watermark, no signature.
Avoid: daylight forest, desert, sand surfing, moonlit lake, outdoor ruins, upright pose, running, jumping, adult body, sexualization, cleavage, duplicate character, extra limbs, extra fingers, fused hands, cropped head, cropped feet, photorealism, 3D render, flat front-facing composition, blur.

結果4:月夜の蓮で片足ダンス

静かな差別化
巨大な満月の下、光る蓮の上で片足のダンスをするナヒーダと水面反射のAI生成イラスト

月夜・水上ダンスプロンプト

Use case: stylized-concept
Asset type: vertical social-media anime illustration and blog experiment result, fourth image in a five-scene series
Primary request: Create a premium Japanese 2D anime key visual of Nahida from Genshin Impact, faithfully recognizable with her small childlike proportions, pale silver-white bobbed hair with soft green-tinted ends, long pointed elf-like ears, bright green leaf-shaped eyes, and ornate white-and-green Dendro dress with gold and leaf motifs. Keep the character wholesome, nonsexualized, and age-appropriate.
Scene/backdrop: a glass-still lotus lake beneath a huge full moon at blue hour, distant Sumeru mountains in mist, sparse glowing lotuses, stars reflected in dark water, thin rings of light spreading from the center.
Subject/action: Nahida performs a gentle one-foot ballet balance on the center of a single enormous luminous lotus. One leg lifts behind in a graceful arabesque, both arms open softly, chin turned toward the moon, and her reflection is clearly visible below. Small Dendro fireflies form a loose halo rather than a violent effect. Her expression is peaceful and quietly joyful.
Style/medium: polished commercial Japanese anime illustration, crisp delicate line art, refined cel shading, luminous water reflections, elegant atmospheric depth, clean readable character design.
Composition/framing: 4:5 portrait, full body visible, distant frontal wide shot with near-symmetrical composition, long vertical reflection, large negative space around the figure, no tilted horizon, no extreme foreshortening, face readable but not a close-up.
Lighting/mood: cool moonlight, soft mint rim light, silver-blue reflections, serene, sacred, dreamlike.
Color palette: midnight blue, moon silver, mint green, pearl white, faint lavender.
Constraints: only one Nahida plus her natural water reflection; accurate childlike facial and body proportions; natural hands and limbs; clearly different from forest, desert, and library scenes; no text, no logo, no watermark, no signature.
Avoid: falling, surfing, upside-down pose, indoor architecture, desert, forest canopy, fiery sunlight, aggressive action, adult body, sexualization, cleavage, duplicated real character, extra limbs, extra fingers, fused hands, cropped head, cropped feet, photorealism, 3D render, cluttered background, blur.

結果5:遺跡守衛の手と朝焼け

親しみ・スケール
朝焼けの遺跡で巨大な遺跡守衛の手の上にしゃがみ、手を振るナヒーダのAI生成イラスト

遺跡・しゃがみプロンプト

Use case: stylized-concept
Asset type: vertical social-media anime illustration and blog experiment result, fifth image in a five-scene series
Primary request: Create a premium Japanese 2D anime key visual of Nahida from Genshin Impact, faithfully recognizable with her small childlike proportions, pale silver-white bobbed hair with soft green-tinted ends, long pointed elf-like ears, bright green leaf-shaped eyes, and ornate white-and-green Dendro dress with gold and leaf motifs. Keep the character wholesome, nonsexualized, and age-appropriate.
Scene/backdrop: ancient overgrown Sumeru ruins at sunrise, broken stone arches covered in moss and tiny flowers, a dormant colossal Ruin Guard half-buried in the valley, warm fog and distant birds crossing the light.
Subject/action: Nahida crouches lightly on the enormous open stone-and-metal palm of the dormant Ruin Guard. One knee is raised, one hand rests naturally on the machine finger for balance, and her other hand gives a small playful wave toward the camera. Fresh vines and white flowers awaken along the machine beneath her. Her expression is bright, mischievous, and inviting.
Style/medium: polished commercial Japanese anime illustration, crisp controlled line art, refined cel shading, detailed weathered machinery and vegetation, cinematic depth, clean facial readability.
Composition/framing: 4:5 portrait, medium full-body framing from a close low camera placed near the machine palm, giant fingers create foreground leading lines, Nahida sits in the upper-middle third, face larger than in the lake scene, strong scale contrast between the tiny character and colossal machine, no circular magic frame.
Lighting/mood: warm coral sunrise, soft mint bounce light from new growth, hopeful, playful, exploratory.
Color palette: coral gold, weathered bronze, moss green, pearl white, soft sky cyan.
Constraints: only one Nahida; the Ruin Guard is dormant and nonviolent; accurate childlike facial and body proportions; natural hands and limbs; clearly different from forest free-fall, desert surfing, upside-down library, and moonlit dance; no text, no logo, no watermark, no signature.
Avoid: combat, weapon, falling, surfing, upside-down pose, ballet pose, dark library, moonlit lake, desert dunes, dense forest canopy, adult body, sexualization, cleavage, duplicate character, extra limbs, extra fingers, fused hands, cropped head, cropped feet, photorealism, 3D render, blur, clutter hiding the face.

95枚を並べて分かった差別化の優先順位

背景名だけを変えても、人物が中央で直立したままなら印象は似ます。今回もっとも効いたのは姿勢の軸です。頭側から落ちる、横へ滑る、逆さに浮く、片足で立つ、低くしゃがむという身体の大形状を先に分け、その後でカメラと光を割り当てると、生成結果のシルエットが重なりにくくなります。

  • 最初に5つの身体シルエットを言葉で分ける
  • 各案でカメラの高さと向きを重複させない
  • 朝・天気雨・真夜中・月光・日の出のように光源を変える
  • 直線、S字、同心円、左右対称、巨大前景で視線経路を変える
  • 感情も冒険、集中、好奇心、静けさ、親しみへ分ける
  • 共通キャラ条件は毎回同じ文章で固定する

10別キャラへ応用するときのプロンプト設計

キャラクターを変更する場合も、5案すべてを一度に書き始める必要はありません。最初に『キャラ固定文』を作り、その下へシーン、動作、カメラ、光、配色、案固有のAvoidを追加します。共通部分と差分を分離すると、衣装精度を保ちながら大胆な場面変更を試しやすくなります。


11結論:ランダム生成ではなく、差分を先に設計する

ナヒーダの識別要素だけを固定し、姿勢、背景、画角、時間、感情を先に分けることで、5枚は同じキャラでも別企画として読める結果になりました。数を増やす前に差分表を作る方が、似た画像を後から捨てるより効率的です。次回も人気候補からキャラを選び、この5軸が重ならないプロンプトを作れば、記事とSNSの両方で使えるシリーズを継続できます。

よくある質問

Q同じキャラのAI画像が似た構図になるのを防ぐには?

背景名より先に、落下、横滑り、逆さ、片足立ち、しゃがみのような身体シルエットを分けてください。その後にカメラの高さ、時間帯、主な視線誘導を各案へ割り当てると重複が減ります。

Qなぜ今回は同じプロンプトを5回使わなかったのですか?

同じプロンプトの複数生成は安定性を見る検証に向きますが、ポーズや背景の差別化には限界があります。今回は投稿シリーズとして明確な違いを作ることが目的なので、キャラ固定文だけを共有し、シーン部分を5種類に分けました。

QSNS向けに最も使いやすいナヒーダ画像はどれですか?

結果2の砂漠サーフィンは速度感、暖色と緑の対比、顔の読みやすさが揃っています。意外性なら結果3、静かな美しさなら結果4、親しみなら結果5が向きます。

Qナヒーダらしさを保つプロンプトの要点は?

銀白色のボブ、緑の毛先、尖った耳、葉形の緑の瞳、白・緑・金の草元素衣装を共通文として毎回同じ位置に書きます。許可された参照画像を使える場合は、さらに衣装精度を高められます。

Q子どもらしいキャラクターを安全に生成するには?

健全、非性的、年齢にふさわしい、小柄で幼い体格という条件を肯定文に入れ、ネガティブ側でも大人の体格、性的なポーズ、露出、胸の強調を除外します。

この記事のタグ

この記事をシェア