Remember when AI image generators couldn’t keep a character’s face straight between two pictures? Well, Google just dropped its much-hyped new model (codename: nano-banana 🍌), now known as Gemini 2.5 Flash Image, that you can try for free on AI Studio.
The big deal? This model nails character consistency like nothing we’ve seen before. You can place the same character in a desert, then underwater, then at a disco — and they’ll actually look like the same person. Wild.
Here’s what happened when we tried to use ChatGPT Image to edit my face for The Neuron podcast’s YouTube thumbnails:
Even if you’ve never watched our podcast, you can probably tell that’s not at all what Grant looks like… they’re not even consistent among themselves! Though TBH, that first guy is pretty cute.
What makes nano-banana special
- Character consistency that actually works: Google built a template app showing how you can keep characters looking identical across scenes.
- Edit photos or drawings with just words: The photo-editing demo lets you remove people, blur backgrounds, or colorize photos using natural language…and this co-drawing demo lets you draw and ask AI to fix it.
- Actual world knowledge: Unlike other image models, this one knows stuff — like how the co-drawing demo turns doodles into learning experiences.
- Multi-image fusion: You can now merge multiple images; for example, you can drag and drop objects between images seamlessly with their home canvas template.
At $0.039 per image (just round that up to $0.04 per image lol, or $40 per 1K images), it’s surprisingly affordable for what you get.
The speed at which this is spreading is nuts
OpenRouter just added it as its first-ever image model (out of 480+ models!), fal.ai is bringing it to its developer community, and Adobe just integrated it into Firefly.
That’s right — Adobe is now offering both its own Firefly models and Google’s Gemini, plus models from OpenAI, Black Forest Labs (which makes FLUX), Runway, Pika, Ideogram, LumaAI, and others. It’s like the AI model wars just turned into an AI model peace treaty.
Why this matters
We’re watching the shift from “my model is better than yours” to “here’s a buffet of models — pick what works.” Adobe gets it. Instead of forcing creators into one ecosystem, they’re becoming the Switzerland of AI creativity.
We’re not sure if this has something to do with the big labs giving up the ghost on the “one model to rule them all,” scale-pilled AGI-at-all costs mode of marketing AI, or just succumbing to the reality that people just want the best model that works for them, but it’s an interesting development.
Either way, it’s becoming increasingly apparent that the future of AI isn’t about one giant model to rule them all — it’s about having the right tool for each unique job.
Editor’s note: This content originally ran in our sister publication, The Neuron. To read more from The Neuron, sign up for its newsletter here.