The Rundown:

In this guide you’ll see the key similarities and differences between Google’s Nano Banana Pro and ChatGPT’s new image generation model. You’ll also learn how to build your own comparison matrix that you can reuse for other model comparisons.

We also tested Google’s advanced image generation tool (AI Studio) to see if it produces better results than just using Nano Banana Pro.

💡Pro tip: If you don’t have Gemini Pro you can use Nano Banana Pro for free here. Try opening an incognito browser if you have trouble.

Who is this useful for:

• Business owners who want to automate asset creation
• Graphic designers using AI for designs
• Anyone that wants to generate better AI photos

What You Will Build:

What You Need to Get Started:

• Gemini Pro account
• Free or paid ChatGPT account
• (optional) Google Workspace account with billing enabled for AI Studio
• (optional) This Notion template for your rating matrix

Step 1

Decide on your use cases. We picked five:

• Logo for a taco truck
• Website graphic for a couch manufacturer
• Social post for an energy drink brand
• Marketing materials for a travel agency
• Arbitrary prompt of something that would be easy to tell if it’s “fake”

Outline your testing rules.

Here’s ours:

1. Same prompt for each model
2. 4 images generated per model
3. Models will be graded 1-5 on
4. 1. Prompt consistency: How well did it follow the prompt across variations?
   2. Creativity: Was each variation noticeably different, subtly different or carbon copies?
   3. Utility: Could you use one or more of the variations for the actual use case?
   4. Quality: Which images look the least like AI slop?

Step 2

Feed your use cases into Claude/ChatGPT/Gemini and ask it to generate prompts structured as json for each.

”Here’s my use cases: [use cases] write me a json prompt for each of the 5 use cases. invent a different, fun brand for each use case and for arbitrary image do something detailed/easy to see if it's fake like a crowd of sports at an nba game. Include instructions to generate 4 variations per prompt in a 4x4 grid.”

If you just want to test it on your brand, ask the LLM to interview you about your brand then write prompts for each use case.

💡Pro tip: To save on time/tokens, tell the LLM to write a prompt that will generate 4 variations in a 4x4 grid.

Step 3

Create a scoring matrix that looks like this. If you’re lazy, duplicate our notion guide here.

| Use Case | Model | Consistency | Creativity | Utility | Quality | Overall Rating |

|-------------|----------------------------|-------------|------------|---------|---------|----------------|

| Use case 1 | Nano Banana (Gemini) | | | | | |

| Use case 2 | ChatGPT Image | | | | | |

| Use case 3 | Nano Banana Pro (Vertex) | | | | | |

Step 4

• Navigate to Gemini. Click the “Create image” button to add Nano Banana

• Navigate to ChatGPT/images
• Optional: Open Nano Banana Pro in AI Studio. Click the “Playground” link on the left side to open new chats.

Step 5

Feed each use case prompt into each tool and hit generate.

Open a new chat for each use case so you clear context.

Rate the outputs based on each criteria.

Going Further:

• Try out a few of these use cases with your brand in both ChatGPT image and Nano Banana Pro.
• Start by asking ChatGPT or Gemini what the task is and to ask you 3-5 questions in an interview. It will use the answers to write 1-3 image prompt variations. Then use the one you like best.
• Typically you should ask for variations in a 4 x 4 grid. When you find one you like, you can ask the LLM to generate that one by itself.