As you probably know by now, Google officially revealed that they are the official source of the mysterious “Nano Banana” that had been tearing up the LLM leaderboards. I spent so much time reading the blog posts and watching the YouTube videos that I missed my Wednesday newsletter entirely. So if you just need the TLDR; yes Nano Banana (or Gemini 2.5 Flash something or other…) is a really big deal. It’s not perfect, but’s a giant leap forward in what we’ll call “context based image editing”. This means you can continue to describe what you want, or make edits with an ongoing chat conversation. It’s certainly not new, as ChatGPT has had this ability since March and Flux Kontext added in May. It may not even be the best, but we’re still discovering different uses cases every day.
It’s Really Smart
What’s really special here is it seems to have more reasoning baked in, it seems to have more cultural context than just an image generator. For example, here I asked for the banana from the Velvet Underground by Andy Warhol to be placed on my shirt.
It’s NOT a Photoshop Killer
Every tech bro on YouTube was shouting “Photoshop Killer” and Adobe’s stock did tank a little, but it does not replace Photoshop. It just raises the bar for everyone and expands or understanding of what can be done with an LLM and a single image. I watched several demos this week where the creator started with a single image of a unique character, and then turned that into 6 or 8 different camera angles or scenes. Each of these were then turned into 8 seconds of video and you very quickly have a minute of storytelling.
It’s Great at Combining Objects
I watched several demos of multiple items of clothing being “dressed” onto an existing image of a model and they were almost flawless. Like in this Weavy demo, the women in black wears the sunglasses and outfit of the woman in the middle; final result on the right.
It Understands the New Environment
Objects almost always fit in the right place and seem to understand scale, light, and the laws of physics.
It Can Convert Your Sketches Into Images
Let’s pretend you cannot even draw, well that’s not a stretch for me, but you still want to sketch out your ideas.
It Handles Product Details Incredibly Well.
In previous context based image editors, we expected our fine typographic details to get warped or distorted. Nano Banana appears to preserve every detail, even thin type tilted on an angle
Is it the best? Well it has some strengths and weaknesses. Occasionally other models outshine it, but I love using it and I feel like I can do almost anything with it.
Negatives: It’s limited to a square aspect ratio, image quality is a little low and degrades over time (more prompts).
Where: It’s suddenly on every image generation platform (Freepik, OpenArt, Krea, etc), even Adobe Firefly. You can try it for free on Gemini.
I have barely scratched the surface of what it can do. I’ll be back after Labor Day with a more detailed report. Or you can watch my Weavy demo.
That’s it for today! Have a great Labor Day Weekend! I’ll try to pull myself away from the computer and go outside for a few hours.