AI Image Generation: From Nano Banana to Fal.ai

Text in. Image out. Pennies per print.

Image generation has gotten absurdly good. But the tools can be confusing. Here's how we think about it and what we actually use.

The Model Provider Approach

Instead of subscribing to individual AI image services, we use Fal.ai as a model provider. Fal.ai gives you access to dozens of image models -- including Nano Banana -- at raw API cost.

This means instead of paying $20/month for one service, you pay fractions of a cent per image across whatever models work best for your needs.

Text-to-Image vs. Image Editing

There are two fundamentally different operations, and mixing them up will frustrate you:

Text-to-image: You describe something from scratch. "A modern coffee shop interior with warm lighting and exposed brick." The AI creates an entirely new image.

Image editing: You start with an existing image and modify it. "Take this photo and change the wall color to navy blue." The AI preserves the original and adjusts specific elements.

Two fundamentally different operations. Know which one you need.

The control in image editing is what makes Nano Banana special. You're not hoping the AI understands your vision -- you're directing it frame by frame.

Our Workflow

1. Define the style: We often start by asking Gemini about specific cinematographic or design styles. "Describe the visual aesthetic of A24 films" gives us a vocabulary to feed into image generation.
2. Write the prompt: Using that style vocabulary, we craft a detailed prompt.
3. Generate via Fal.ai: Send the prompt through Fal.ai to whichever model fits.
4. Refine with editing: Take the best output and use image editing to adjust details.

Research. Prompt. Generate. Refine. Repeat.

This process produces consistent, brand-aligned images across entire websites and campaigns.

Integrating with Your Codebase

Inside Cursor, we can generate images using MCP connections to Fal.ai. We used this exact workflow when cloning and redesigning a client website in a day. The AI maintains context from our project -- style guides, brand colors, existing images -- so new generations stay consistent with what's already been built.