Using Nano Banana for Image Editing

On By Kate ScottIn Image Generator, LLMs, Tools, Use Case

There’s been a lot of chat and excitement about Gemini’s new image generator – Nano Banana (where do they get these names from?). It’s even available in Adobe Photoshop now.

While it’s a good image generator, its real strength is in editing. You can edit via text prompt and unlike previous models from various platforms, this one makes ONLY the changes you ask for. It doesn’t re-generate (and change – subtly or otherwise) the entire image.

As an example, I wanted to come up with a picture for an Instagram post. I’d done a guest post on a writing blog and needed something that was both eye-catching and sort of made sense with the text I wanted to add.

I used ChatGPT to suggest some social snippets from the post, then got it to create an image prompt for this one: Fairytales give me scaffolding. A flexible structure I can build on, twist, or completely subvert when my characters decide to run off and solve crimes with cats instead of going to the ball.

I could have used ChatGPT’s built-in image generator, but they only allow three images a day. Gemini doesn’t have a set restriction. Plus, ChatGPT tends to make other little changes you didn’t ask for.

Surreal illustration of a writer’s desk turning into a winding staircase made of paper, branching into unexpected directions, soft light, magical realism, textured illustration.
Chat GPT Prompt: Surreal illustration of a writer’s desk turning into a winding staircase made of paper, branching into unexpected directions, soft light, magical realism, textured illustration.
Paper staircase on forest background
Gemini editing prompt: turn the background into a soft fairytale forest backdrop, while keeping the floating elements
Put a small, black cat sitting, licking one of its paws, in the bottom right corner
Second editing prompt: Put a small, black cat sitting, licking one of its paws, in the bottom right corner
Nano Banana image with text snippet added
Final with text added in Canva (any AI image generator is likely to struggle with that much text)

You can also upload and change images generated in other tools. Which, if you need something in an aspect ratio other than square, is helpful. The below picture of the girls is from Ideogram, I uploaded it and got Gemini to change them to boys.

Mirror type pic of two identical young women, a happy hockey player and a serious tennis player
Ideogram prompt: An image of two identical girls, facing each other like mirror reflections. one girl is a muddy, happy field hockey player, with a hockey game in her background. the other, with the same colouring and features but a serious expression is in tennis gear with a fancy club tennis court behind her.
The image of the young women, edited to young men
Nano Banana editing prompt: please turn the two girls into boys with short black hair

Pity they both turned serious but, I could always do another edit.

If you want to generate images in Gemini, you need to select the image generator as one of the tools in the prompt box (it’s helpfully marked with a banana).

Gemini Prompt box tools list

And, yes, you can use it to make deepfakes… although when you upload an image, Google does ask you to confirm you have permission to use whatever it is you’re uploading.

Gemini Image Permissions

Since I’m allowed to use pictures of me… I popped in a pic of myself in, and sent myself on holiday.

Me
Me magically transported to a tropical beach

Is it flattering? No. Is recognisably me, in a completely different place, pose and outfit? Oh yeah.

The two takeaways:

  1. Nano banana is a super easy-to-use tool that creates and edits pictures for free, and
  2. Don’t believe anything you see on social media (or anywhere else, frankly).

Leave a comment