Published: 2026-05-19
Keywords: nanobanana prompts, fashion magazine cover, cinematic portrait, Nano Banana Pro, director-level prompts
90% of people still rely on “tag-style prompts” and hope for luck — I used this director-level prompt once and got an 8K cinematic portrait ready for print.
Since Google DeepMind launched Nano Banana Pro (Gemini 3 Pro Image), AI image generation moved from “aesthetic collage” to reason first, render second — physics-level understanding.
What truly shocked me wasn’t how beautiful it could draw, but that it could follow 100% of my half-screen complex instructions — including physics like blurred pedestrians in the foreground under long exposure while the subject stays razor sharp.
Today I’m publishing my go-to ultra-realistic cinematic fashion portrait prompt in full, and breaking down exactly how it unlocks every strength of Nano Banana Pro. Copy, paste, tweak — and ship.
1. What a production-ready prompt looks like (copy & reuse)
Below is the prompt I built for “the model in reference image 1.” Clothing and scene are placeholders — everything else can be used as-is.
How to use: In Google AI Studio or Gemini App (Thinking mode on), select Nano Banana Pro, upload a clear front-facing reference (face unobstructed), paste the prompt below, replace Clothing and Scene/Environment, then generate.
Note: The prompt below is the original Chinese version — copy it unchanged for best results with Nano Banana Pro.
主题设定：超逼真的电影感肖像，高端时尚，8K质感，画幅杂志风格摄影。
主体描述：仅使用上传的参考图1作为主要拍摄对象，面部特征、骨骼结构和自然肤质（可见毛孔）需100%匹配，妆容与参考图角色保持一致。不得改变种族或性别特征。
发型描述：与参考图1人物的发型和发色保持高度一致，发丝被风吹拂，飘散在脸上，以时尚的方式部分遮挡面部特征。
妆容设定：
1 底妆：通透干净，轻薄雾面带微光
2 腮红：粉色集中在脸颊与鼻梁两侧
3 眼妆：自然放大，睫毛纤长，眼神清澈
4 唇妆：水润嘟嘟唇，淡粉到蜜桃渐变，唇峰柔和
超逼真的皮肤纹理，毛孔清晰可见，细节自然，纹理细腻。
服装设定：【请填写你的服装描述，例如：黑色极简剪裁羊毛大衣 + 哑光皮裤】
场景/环境：【请填写你的场景描述，例如：雨后黄昏的东京表参道，地面倒映暖色灯光】
动作/动态：前景有穿着时尚的行人快速移动，动态模糊（长曝光效果）挡住部分画面和镜头。拍摄对象保持完美清晰、沉静。拍摄对象静止，轻微仰头，凝视镜头（时尚杂志模特的pose），位于画面的右侧。
构图：低角度仰拍人物，全身横幅肖像，主体人物位置偏右构图，线条简洁流畅。
相机：24mm广角镜头视角，浅景深，电影感虚化效果，专业的时尚写实风格。
光照：柔和的电影感光照，暖色调，营造氛围感。高光控制得当，阴影保留细节。
色彩/后期：高细节，高级色彩分级，自然的肤质纹理，微妙的胶片感（无明显颗粒）。无文字、无徽标、无水印。
画面比例：16:9（横幅全身），2K HD输出。
Usage steps
- In Google AI Studio or Gemini App (Thinking mode), select Nano Banana Pro
- Upload a clear reference image 1 (front-facing, face unobstructed)
- Paste the prompt and replace Clothing and Scene/Environment with your ideas
- Generate — one pass for identity-consistent, makeup-accurate, motion-blur-aware results
2. Why this prompt makes Nano Banana Pro over-deliver
Ordinary prompts “ask for things.” This one is a shot list. Six design choices hit Nano Banana Pro’s strongest capabilities.
1. Strict reference constraints → 100% identity & makeup
“Use only uploaded reference image 1… facial structure and pore texture 100% match”
→ Nano Banana Pro supports up to 14 reference images and keeps the same character across generations. Stating “do not change race/gender” triggers logical checks — no generic “beauty filter” face.
2. Long-exposure foreground blur + still subject → physics-level motion
“Fashion pedestrians move fast in foreground with motion blur; subject stays perfectly sharp”
→ Most models draw “blurry blobs.” Nano Banana Pro reasons: faster + closer = stronger blur; locked focus on a still subject = sharp. That’s “think before you draw.”
3. Makeup broken into steps → micro-detail control
Base, blush, eyes, lips — each with placement notes like blush on nose bridge.
→ The model maps aesthetic instructions to pixel regions — no lipstick bleeding or full-face blush disasters.
4. Pro photography parameters → built-in camera knowledge
“24mm wide low angle + shallow DOF + warm cinematic light”
→ It knows how 24mm stretches perspective and how low angles amplify presence — you don’t need to be a photographer.
5. Clean negative constraints → no junk in frame
“No text, logos, watermarks, heavy grain”
→ Many models add fake watermarks or timestamps. Explicit negatives become reasoning constraints for clean commercial output.
6. Full natural language + numbered structure → efficient parsing
Complete sentences and numbered lists — not “8k, cinematic, masterpiece” tags.
→ Tag spam is a downgrade for a multimodal reasoning model. Natural language + logic chains conditions instead of random sampling.
3. Nano Banana Pro vs traditional diffusion (why switch)
| Capability | Traditional diffusion | Nano Banana Pro |
|---|---|---|
| Multi-step complex prompts | Often drops after item 3 | Fulfills in order; understands “keep X while Y” |
| Character consistency | Heavy rerolls + face fix | One reference; multiple outputs stay on-model |
| Text rendering | Garbled CJK/Latin | >95% Chinese accuracy; handwriting supported |
| Long exposure / motion blur | Blur blobs | Understands speed, distance, depth of field |
| Physics | Floating objects, wrong shadows | Light simulation (reflections, shadows) |
| Resolution | Mostly 1K–2K | Native 4K — print-ready detail |
In short: other models paint what they see; Nano Banana Pro understands, then draws.
4. Beyond portraits: top use cases
Besides fashion, “long context + physical reasoning” also excels at:
- E-commerce product shots: casual product photo + “45° side view, soft rim light, shadow on beige backdrop” → Amazon-grade set
- UI/UX mockups: structured prompts for brand manuals, dashboards — layout logic holds
- Technical diagrams: upload a paper PDF → Transformer architecture or infographics with clear labels
- Comics & storyboards: multi-character identity across panels
- Education: handwritten geometry proofs with correct auxiliary lines and steps
5. Quick start in three steps
Step 1: Choose your platform
- Free: Google AI Studio → “Nano Banana Pro” or
gemini-3-pro-image-preview - App: Gemini → Tools → Create image → Thinking mode
Step 2: Reference + structured prompt
Swap character, outfit, scene, and action placeholders. Keep makeup detail, camera/lighting specs, and negatives (no text/watermark).
Step 3: Output & iterate
- Start with 1K preview, then ask for 4K when happy
- Refine in chat: e.g. “Reduce foreground blur; keep subject sharp”
6. Final word
AI won’t replace photographers — but people who think like directors will replace people who only write tags.
Nano Banana Pro isn’t magic “one-click blockbuster.” It’s a crew that reads shot lists. Write clear, ordered, physics-aware prompts — and it renders once.
Stop gambling on “pretty girl, HD, cinematic light.” Write your own director’s notes and let AI execute.