最近这段时间谷歌Nano Banana太火了,基本上改变了图像编辑领域的范式,但是很多小伙伴不知道有哪些比较好用的提示词。

下面是收集到的比较推荐的提示词,主要是英文,大家也可以结合翻译工具换成中文。

(总结自官方/媒体):
- 关键强项是多轮细化与主体一致性(迭代式编辑、文字排版也OK)。(Google Developers Blog, Google AI for Developers)
- 支持把同一角色放入不同场景、两图融合等;但有些基础编辑(如精确裁剪)仍有限。(Google AI Studio, Axios, PC Gamer)
- 做“置换环境”的写实合成时,场景一致性与光源匹配更关键(社区经验)。(Reddit)
一.分镜 / 角色一致性(Storyboard)
适合“上传人物/虚拟角色 → 直接出分镜图”。
1. Turn the uploaded portrait into a collectible **character figure**. Place a branded box behind it (box shows the character image), a computer screen on the side showing a Blender modeling viewport, and a round plastic base under the figure. **PVC look**, indoor studio, **realistic lighting, film tone**, keep face identity.
2. Keep the same character. Scene 1: **subway platform**, cool blue ambience, rim light from behind; camera at chest height, medium shot.
3. Same character, **coffee shop close-up**, shallow depth-of-field (bokeh), warm golden hour light, hands holding a paper cup with subtle steam, keep identity and hairstyle.
4. Same character, **office night scene**, glass reflections, city lights bokeh, three-point lighting, medium-wide.
5. Same character, warehouse with volumetric light rays through windows; dust motes, slightly desaturated cinematic look.
6. Storyboard 6 frames: hero intro, establishing shot, close-up on hands, over-the-shoulder UI shot, product hero, end card. Keep character identity across frames, consistent wardrobe.
7. Turn pose to a walking forward action, slight motion blur on legs, maintain facial details.
8. Same character in **rainy street**, wet asphalt reflections, neon signs; keep eye color and freckles.
9. Same character, studio packshot background in neutral gray; soft top light, fill card from camera left; expressions neutral.
10. Generate a shot list of 8 frames for an unboxing: box on table → hands open → figure removed → detail close-ups → turntable shot → closing hero. Keep product scale and branding.
二.电商 / 产品合成(商品+场景融合)
1. Merge image1 (product) with image2 (modern kitchen). Place the product on a **marble countertop**, keep realistic reflections and shadows; color temperature ~5000K, slight vignette.
2. Cosmetics bottle on **acrylic pedestal**, seamless paper backdrop, softbox from right, subtle gradient background, specular highlights controlled.
3. Sneakers on **concrete block**, low-angle hero shot, studio rim light, faint haze for depth.
4. Smartphone **floating exploded view**: separated layers for screen, mid-frame, camera module; dark studio, glossy reflections managed.
5. Watch macro shot, **90° top-down**, soft diffused light, brushed metal texture visible, anti-alias text on dial.
6. A/B variants: same layout, change only background material (wood / stainless / terrazzo), keep product color constant.
7. Lifestyle placement: living room sofa, daylight from window camera left, product on side-table; white balance matched to scene.
8. Promo trio: front / 45° / top angles in one session; consistent brand palette and shadows.
三.人像妆造 / 服饰替换(注意保持身份一致)
1. Keep face identity; change hairstyle to **retro bob**, add thin eyeliner, soft skin retouch while preserving pores, realistic hair strand flyaways.
2. Outfit swap to **streetwear (oversized hoodie + cargo pants)**, neutral studio background, soft Rembrandt pattern lighting.
3. Add sunglasses with subtle reflections, ensure eyes still faintly visible through lens tint.
4. Formal portrait, **key light 45°**, fill −1EV, background seamless #EAEAEA, color-accurate skin tone.
5. Outdoor candid, overcast sky as giant softbox, slight film grain, maintain freckles and moles.
6. Headshot for ID: straight-on camera, no smiles, flat even light, neutral background, maintain exact facial geometry.
四.光影与摄影语汇(直接“像摄影师”说话)
1. Film noir: **hard key from high angle**, deep shadows, practical lamp in background, smoke for volume.
2. Golden hour backlight, **rim light on hair**, flare streaks, warm highlights ~5600K.
3. Top-down softbox with negative fill on both sides to shape the face, subtle catchlight.
4. High-key beauty, near-white background, contrast low, speculars controlled.
5. Rain reflections, **wet-down street**, neon glow; long-exposure light trails.
6. Product hero: **flagged highlights**, gradient card for polished metal, reflection on base kept tight.
五.材质与质感(让“质地”更可信)
1. PVC plastic look, slightly translucent edges, smooth specular roll-off, no micro-scratches.
2. Anodized aluminum, fine linear brushing, soft anisotropic highlights.
3. Matte ceramic, soft diffuse, micro-pitted texture visible only in macro.
4. Brushed stainless steel with subtle fingerprints near button, realistic.
5. Frosted glass bottle, internal light scatter, bright meniscus line on liquid.
六.文字 / 海报(含文本渲染)
官方强调对包含文字的图像有针对性优化,适合海报/Logo/封面。(Google Developers Blog)
1. Poster: bold headline “AI VIDEO STUDIO”, sub-headline small caps, clean sans serif, centered layout; hero product bottom-right, strong grid, optical margin alignment.
2. Minimalist **event flyer**: date big numbers “09·10”, venue in small line below; keep text crisp and aligned.
3. App mock poster: phone frame with screen UI, headline 5-6 words, badge “NEW”.
4. Brand logo lockup on dark background, subtle glow, high legibility, no distortions.
5. YouTube thumbnail style: big face close-up + 3-5 word punchy title; high contrast; rule of thirds.
七.合成 / 场景替换(两图融合与换景)
官方/媒体均提到多步编辑与两图融合是亮点;注意让光源/焦距/色温匹配以更写实。(Axios, Google AI Studio, Reddit)
1. Replace background with snowy mountain scene; match key light from camera right; footprints faint in snow.
2. Place subject into **museum gallery**, polished floor reflections, soft track lighting from ceiling.
3. Combine portrait with **night city rooftop**; wind-blown hair; cool key, warm practicals behind.
4. Merge product with **bathroom marble vanity**; soft window light, mirror reflection consistent.
5. Subject into **library**, warm tungsten practicals; lens fall-off consistent with 50mm look.
6. Remove clutter from desk; keep laptop and mug; fill missing areas with plausible textures.
八.镜头语言 / 构图(给“画面调度”可控性)
1. 35mm equivalent, eye-level, medium shot, rule of thirds.
2. 85mm portrait, shallow DOF, background separation.
3. Top-down **flat-lay**, grid-aligned objects, neat spacing.
4. Low-angle hero shot, leading lines converge to subject.
5. Tilt-shift miniature look, thin plane of focus through subject.
九.特效 / 情绪氛围
1. Light fog + god rays from window; particles visible only against backlight.
2. Subtle film grain (fine), chromatic aberration minimal, letterboxed.
3. Rain streaks + surface ripples on puddles, reflections intact.
4. Fireflies/bokeh spheres floating, depth-based size variation.
5. Confetti burst, shallow DOF; motion blur only on confetti.
十.多轮对话脚本
多轮迭代是官方推荐的最佳实践。(Google Developers Blog)
脚本 A|人物置景分镜(3 步)
- 第 1 轮:Upload the portrait. Place the same person in a sunlit café terrace; warm golden hour, shallow DOF, cinematic.
- 第 2 轮:Keep identity. Add a laptop with code editor on screen; add a cup with visible steam; adjust white balance slightly cooler.
- 第 3 轮:Tighten composition to medium shot; brighten catchlights; add gentle film grain.
脚本 B|商品三视图(3 步)
- 第 1 轮:Front hero shot on neutral backdrop; soft top light, clean shadows.
- 第 2 轮:Add 45° view; match exposure and shadow direction.
- 第 3 轮:Add top view; assemble a 3-panel layout with consistent margins.
脚本 C|角色手办化(3 步)
- 第 1 轮:Turn subject into a PVC figure on round base; indoor studio, film tone.
- 第 2 轮:Add branded box behind + monitor showing Blender modeling.
- 第 3 轮:Refine edges on hair and fingers; ensure specular highlights are smooth.
十一.N8N 集成用的字段模板
{
"prompt":"Turn the uploaded portrait into a collectible character figure on a round plastic base; branded box behind; monitor shows Blender viewport; PVC look; indoor studio; realistic lighting; film tone; keep identity.",
"style_notes":"cinematic, shallow DOF, subtle film grain",
"consistency_notes":"keep face geometry, eye color, hairstyle",
"scene_alt":["sunlit café terrace","office at night","rainy neon street"],
"product_merge":{
"bg":"modern kitchen marble countertop",
"shadows":"match key from camera right",
"reflections":"controlled, realistic"
}
}
实操要点(精炼)
- 用段落描述场景,不要只堆关键词;先大结构,再补光影/材质细节(官方写法建议)。(Google AI for Developers)
- 多轮微调(光源方向、景深、裁边、高光控制),每次只改 1–2 个因素(官方推荐)。(Google Developers Blog)
- 角色一致性:上传清晰参考图;固定发型/肤色/面部细节用语,减少大幅度冲突改动(场景/光源匹配更真实)。(Google AI Studio, Reddit)
- 限制认知:有些基础编辑(如精准裁剪到特定比例)体验一般,可用“生成时就指定构图+留白”来规避。(PC Gamer)