原文
Nano Banana · en
{
"render_goal": "Ultra-cinematic, hyper-realistic rainy street portrait with shallow depth of field",
"scene": {
"location": "busy city street",
"weather": "steady rain, wet reflective roads, soft mist in the air",
"time": "overcast daytime",
"atmosphere": "moody, cinematic, emotionally rich"
},
"subject": {
"identity": "girl identical to the reference",
"action": "walking forward through a crowded street while holding an umbrella",
"pose": "natural walking posture, relaxed shoulders",
"expression": "calm, confident, slightly intense",
"eye_direction": "looking directly at the camera from a distance",
"outfit": {
"style": "elegant, modern, attractive",
"fit": "well-tailored, flattering silhouette",
"fabric_behavior": "slightly affected by rain, realistic wet textures"
}
},
"composition": {
"camera_distance": "medium to long shot",
"framing": "subject centered with balanced negative space",
"angle": "eye-level, cinematic street photography",
"focus": "razor-sharp focus on the girl only",
"depth_of_field": "extremely shallow"
},
"background": {
"crowd": "dense crowd walking in all directions",
"motion": "crowd fully blurred with motion blur",
"detail_level": "faces and bodies indistinguishable",
"separation": "strong subject-background isolation"
},
"lighting": {
"type": "soft diffused natural light",
"highlights": "subtle rain highlights and reflections",
"color_tone": "cool cinematic color grading with slight contrast"
},
"visual_style": {
"look": "cinematic street photography",
"realism": "ultra-realistic skin, fabric, and rain physics",
"quality": "8K detail, high dynamic range",
"post_processing": "film-like contrast, subtle grain, no oversharpening"
},
"constraints": {
"accuracy": "match reference pose, framing, mood, and focus precisely",
"avoid": [
"cartoon look",
"overexposed highlights",
"extra subjects in focus",
"stylized or fantasy elements"
]
}
}使い方
- パラメータを含む元のプロンプト全体をコピーします。
- 被写体、文字、構図を自分の要件に置き換えます。
- 対応する外部ツールで使用し、利用条件と結果を確認します。
開始前の確認
- 参考画像の要否を確認してください。互換モデルの表示は、そのモデルでの生成実績を意味しません。
- 名前、数値、画像内の文字を確認します。モデルの版で結果は変わります。
