AI 1GPTimage25 2nano banana2 3Grokimagine2 Retro...
Ratio 3:4Text-to-Image
Creator Reference:@CandyTravelight
Original Post

AI 1GPTimage25 2nano banana2 3Grokimagine2 Retro

Copy this Nano Banana Pro text-to-image prompt to create portrait concepts with 35mm film texture and photorealistic detail. Best in 3:4 framing.

Master PromptClick to select all

三家 AI 画出的港风茶餐厅,今天早上你吃什么? 图1:GPT-image-2.5 图2:nano banana2 图3:Grok-imagine2 提示词 Retro Hong Kong cha chaan teng portrait, 90s Hong Kong film still aesthetic, candid documentary vibe. 人物:25 岁成年东方女性,瓜子脸、细眉、深色眼瞳;皮肤白皙,保留细腻毛孔与自然纹理,拒绝塑料磨皮;淡妆,唇色偏红。黑色长发披肩,发尾带自然微卷,几缕碎发垂在脸侧。 表情:眼神落在镜头外左侧,神情慵懒放松,嘴角带一丝若有若无的笑,像午后发呆被抓拍到。 服装:奶白色纯棉短袖衬衫,领口敞开一颗扣子;深棕色高腰灯芯绒阔腿裤;黑色细腰带;珍珠耳环。穿搭干净利落,复古味藏在细节里。 姿态动作:坐在茶餐厅靠窗卡座,身体微侧向窗户,一只手肘撑着桌面、手指轻握一杯冒着热气的丝袜奶茶,另一只手随意托腮。坐姿松弛不端着。 道具:桌上有菠萝油面包、竹蒸笼里的虾饺皇、金属奶茶杯、水牌菜单、木质筷子筒。 环境:前景是略虚的奶茶杯和蒸笼;中景是人物与卡座桌面;远景是茶餐厅内景——绿色马赛克瓷砖墙、红色霓虹灯牌「茶餐厅」、老式吊扇、玻璃柜里的蛋挞蛋黄酥;更远的食客是柔和的虚影。 光线:室内暖黄吊灯做主光,色温约 3200K,裹着一层薄薄的油烟雾气;窗外阴天冷调天光打进来做轮廓光,冷暖对撞。玻璃柜和瓷砖上有柔和高光反射。 相机:50mm 定焦,f/2.0,半身景别,机位与人眼平齐略低,shallow depth of field,主体清晰、前景远景自然虚化,slightly handheld casual framing,film grain,Kodak Gold 胶片��拟,Eastmancolor 色彩。 情绪:怀旧、慢节奏的午后慵懒,像老港片里被定格的一帧。 质量标签:ultra detailed, photorealistic, cinematic, high resolution, natural skin texture。 负面约束:No oil painting, No anime, No 3D render, No cartoon, no deformed hands, no plastic skin, no oversaturated colors, no text artifacts, no watermark。

Optimal Generation Parameters
Aspect Ratio
3:4
ModelNano Banana Pro
Aesthetic Keynotes35mm Film, Photorealistic, Cinematic, 3D Rendered

Test & Generate with GPT Image 2

Prompt tokens and APIMart GPT Image 2 are pre-loaded. Customize or click generate.

GPT Image 2 生成器

AI 1GPTimage25 2nano banana2 3Grokimagine2 Retro
Nano Banana 2
Creative Blueprint & Recipe Breakdown

How to Master & Adapt This Visual

Professional prompts rely on modular tokens. Use this recipe breakdown to swap subjects, control studio lighting, and eliminate common AI artifacts.

Prompt Variables Blueprint (What to Customize First)

Iterate on one variable group at a time to maintain stylistic consistency:

1Change the subject block first: 三家 AI 画出的港风茶餐厅,今天早上你吃什么?
2Adjust the mood through light and backdrop cues: 5 图2:nano banana2 图3:Grok-imagine2 提示词 Retro Hong Kong cha chaan teng portrait, 90s Hong Kong film still aesthetic,...
3Keep the finish constraints aligned while you test variants: 0,半身景别,机位与人眼平齐略低,shallow depth of field,主体清晰、前景远景自然虚化,slightly handheld casual framing,film grain,Kodak Gold...

3 Style Recipes

Swap the subject, keep the frameRecipe #1

Replace 三家 ai 画出的港风茶餐厅,今天早上你吃什么? with your own subject while leaving the camera and composition cues intact.

Retune the mood, not the structureRecipe #2

Keep the subject block, then rewrite 5 图2:nano banana2 图3:grok-imagine2 提示词 retro hong kong cha chaan teng portrait, 90s hong... to move the image into a new lighting setup.

Change the finish without breaking geometryRecipe #3

Test a new texture or rendering finish by rewriting 0,半身景别,机位与人眼平齐略低,shallow depth of field,主体清晰、前景远景自然虚化,slightly handheld casual... while leaving the scene structure alone.

Production Quality Guide: Best Use vs. Pitfalls

Ideal Use Cases (Where it Shines)
  • •Portrait studies, beauty concepts, and tightly framed editorial references.
  • •Text-to-image ideation when you want a reliable starting frame in 3:4.
  • •Cinematic, mood-first outputs where light direction carries a lot of the final image quality.
  • ✓Use 3:4 on the first run, then test alternate crops after the base composition is working.
  • ✓This prompt suppresses text or logos. Remove that constraint before asking for branded packaging, typography, or visible watermarks.
  • ✓Lock the environment before fine styling changes: 三家 AI 画出的港风茶餐厅,今天早上你吃什么?
Pitfalls to Avoid (Common Mistakes)
  • !Ignoring the recommended 3:4 crop on the first run. The prompt's composition cues were written for that frame.
  • !Changing subject, lighting, and finish together. You lose the stable base that makes the prompt reusable.
  • !Asking for readable type, logos, or packaging text without first removing the no-text constraint.
  • ✕Skip if: You need a wide environment shot or multiple subjects competing for attention in the same frame.
  • ✕Skip if: You need to begin with a radically different crop than 3:4 instead of honoring the existing composition hints.
  • ✕Skip if: You want a loose multi-subject scene instead of a prompt built around one controlled focal point.

Technical Analysis & How This Prompt Works

Copy this Nano Banana Pro text-to-image prompt to create portrait concepts with 35mm film texture and photorealistic detail. Best in 3:4 framing. Ready for...

Keep the base prompt intact for the first run, then iterate on one variable at a time so the composition stays stable. Keep the generator on 3:4 for the closest match. The mood is anchored by 三家 ai 画出的港风茶餐厅,今天早上你吃什么?

Explore Related Curated Categories

More Visual Inspirations

Explore prompts sharing similar aesthetic tokens and composition

View all prompts →