illustration
2:3
Nano Banana Pro Template Design Method
Subject
我用 Cursor + CC（Claude Code）来设计 Nano Banana Pro 的 image prompt 模版。核心思路是把通用的视觉设计原则写入 /CLAUDE.md 作为全局指引，这样在迭代时就不需要反复说明那些共通的方法论——比如「让视觉效果服务于叙事」、「原则优于枚举」等底层逻辑。
💡 从概念到模版的迭代流程
以图 1 的「记忆碎片模版」为例：
有一个初步想法（或者参考图）后，我会直接扔给 CC：「基于这个概念构建新模版」。它会根据全局指引快速生成首版 prompt，然后在 Cursor 里直接跑 API 看效果。 [图 2]
接下来就是视觉校准的循环：看到生成结果后，针对具体问题给反馈——「玻璃平面角度不对，应该始终平行于画布」、「碎片太小了，叙事残缺」。CC 会立即更新 prompt，我确认改进后继续测试。因为全局指引已经覆盖了大量通用原则，这个过程非常高效，不需要每次都重复解释「什么是好的 prompt」。[图 3]
多跑几轮对照测试，验证表现稳定后，就进入最后的精简阶段：检查 prompt 里有没有防御性的冗余描述，仅保留必要的复杂度和设计偏好。
最终归档的模版就是简洁、可迁移的：填充 [Subject] 变量就能复用，AI 会根据语境自动设计。. Try: swap the background to your brand color for a campaign-ready version.
Composition
Lighting
Setting
Style
Details
How to Use This Prompt
Get from copy-paste to final image in 4 quick steps.
1
Copy the prompt
Hit the 📋 Copy button above — your clipboard now has the full prompt, with the tip text auto-stripped.
2
Replace any placeholders
Look for bracketed terms like [SUBJECT], [BRAND COLOR], or {style}. Swap them with your specifics for best results.
3
Pick your aspect ratio
This prompt was tested at 2:3. Square (1:1) works well for social, 9:16 for stories, 16:9 for hero banners.
4
Generate in your favorite model
Paste into Nano Banana 2, GPT Image 2, Midjourney, Flux 2 or DALL-E. Each model interprets it slightly differently — try 2-3 to pick the winner.
Compatible AI Models
This prompt is tested and works well across these models. Click any badge to learn more.
Variation Ideas
Mini tweaks to make this prompt fit YOUR brand or story:
①Try the same composition at three aspect ratios (1:1, 4:5, 16:9) to see which framing tells the story best.
②Swap the main subject for your own product, character or brand identity while keeping the lighting + style intact.
③Switch the lighting (golden hour ↔ studio softbox ↔ blue hour ↔ neon night) to dramatically change the mood.
④Add a single style modifier at the end — e.g. 'shot on 85mm', 'film grain', 'editorial Vogue style', 'cinematic anamorphic'.
Frequently Asked Questions
Which AI model produces the best result for this prompt?
All four major models (Nano Banana 2, GPT Image 2, Midjourney v7, Flux 2) can render this prompt well. Nano Banana 2 excels at character consistency and text. GPT Image 2 is strongest at multi-element scenes with text. Midjourney has the most painterly artistic style. Flux 2 is the best open-source option.
Can I use this prompt for commercial work?
Yes — every prompt on AI Image Free is free for commercial use. The reference image is AI-generated and serves as visual inspiration. The actual image you generate using this prompt is owned by you (subject to your chosen AI model's terms of service).
How do I customize this prompt for my own brand?
Look for bracketed placeholders like [SUBJECT], [COLOR], [BRAND]. Replace those with your specifics. You can also append your own style descriptors at the end (e.g. 'in our brand colors #FF6B35 and #2D3142') for a branded version.
