Professional images with AI

What you'll see in this lesson

1. The map: which model for each job

   No model is best at everything: each one shines at something, and choosing well saves hours of redoing images.

2. Nano Banana 2.1: Google's all-rounder

   It's the image model Google recommends by default: it generates in 1K, 2K or 4K, reasons before drawing and edits turn by turn.

3. gpt-image-2 and OpenAI's new 2.5 family

   Since September 2026 OpenAI offers gpt-image-2.5 in two versions: Flare, the fast one, and Sunburst, the most precise at editing.

4. FLUX, Seedream and Qwen-Image: more options

   For catalogs, for working locally or for text in other alphabets, these three complete the map.

5. Readable text inside the image

   Today's models write short headlines well if you give them the exact text, the position and the typeface.

6. How to write an image prompt

   Describe the image as you'd brief a photographer: shot type, subject, scene, lighting, camera and format.

7. Editing photos and using reference images

   The most useful thing for a business isn't inventing from scratch, but starting from your real photos and changing only what's needed.

8. Usage rights: what you must check

   Before publishing, check three things: who owns the references, whether a real person appears and what the model's license allows.

9. Summary

   Choose the model per job, write the prompt like a brief and check rights and text before publishing.