Everything Nano Banana 2.1 can do

Nano Banana 2.1 is Google's image model for detailed briefs. It sets quoted text exactly, keeps a person the same from picture to picture, counts what you ask for, and can look facts up on the web. It reads up to 14 pictures and 10 videos at once and makes images up to 4K.

Every picture below was made by the step under it. Change a placeholder and press Gen to make your own.

Exact text

To get exact text, put the words in quotes. Here the masthead and the three cover lines are quoted. The masthead sits behind the subject's head because the prompt asks for it.

Generate 7 gens
7 gens

Longer layouts work the same way: quote every line and say where it goes. The next step edits the poster and translates each line into Japanese, keeping the layout.

Generate 3 gens
3 gens
Edit 1 gen
Takes 2
1 gen

Same person, anywhere

Describe a person once in a placeholder, and every step that uses it draws the same person. The cover and this portrait share one. To place them somewhere else, mention the portrait in the next step's prompt.

Generate 2 gens
2 gens
Generate 1 gen
Takes 4
1 gen

Several pictures in one

A step can read more than one picture. Mention each one and give it a role: here the packshot is the product and the second picture is the setting.

Generate 2 gens
Generate 3 gens
Generate 3 gens
Takes 6 7

The same works with four pictures: the portrait, a jacket, a bag and sunglasses. Swap in your own photos.

Generate 1 gen
Takes 4
1 gen

Edit only what you name

Edit changes only what the prompt names. Here it changes the light and the sky, and keeps the person, the pose and the framing.

Edit 2 gens
Takes 5
2 gens

Counts and positions

It follows counts and positions. The prompt asks for exactly five strawberries along the bottom edge. Give each object a number and a place.

If a count comes out wrong, raise Thinking. The price stays the same; it just takes a few seconds longer.

Generate 2 gens
2 gens

A denser layout: nine rooms in a three-by-three grid, each with its own scene, at 4K.

Generate 3 gens
3 gens

Facts from the web

With Search on, the model looks facts up before drawing. The height, year and city in this picture come from the web. Turn it on when the picture has to be accurate.

Generate 2 gens
2 gens

These two steps share one placeholder, the city, so you pick it once. The forecast is looked up when the step runs. Search adds to the price, so leave it off when nothing needs looking up.

Generate 1 gen
1 gen
Generate 1 gen
1 gen

Sizes run from 1:8 to 8:1, up to 4K. For a wide format like 4:1, say what goes in the middle and on each side.

Generate 1 gen
1 gen

To build on any step, press Mix to copy its recipe into your own page.

#guide #nano-banana #prompting

12 views 30 gens