Skip to content
← All articles

On This Day in AI · 25 August 2026

Two Years Ago Google Would Not Draw a Face. Now It Is a Button.

On this day, in 2024, Google quietly started letting its AI draw people again, six months after it had switched the ability off in embarrassment.

August 28, 2024: Google starts letting Gemini make pictures of people again, six months after switching the feature off

August 26, 2025: Google releases Gemini 2.5 Flash Image, nicknamed Nano Banana, and it goes round the world in a fortnight

August 2026: the same thing is a button in Google Photos, and nobody mentions it

On this day, in 2024, Google quietly started letting its AI draw people again, six months after it had switched the ability off in embarrassment.

Back then the whole thing felt fragile. That February, Gemini had been asked for historical pictures and produced people of the wrong era and the wrong background, Google apologised in public, and rather than patch it, took away the ability to generate images of people at all. When it came back that August it came back carefully, to paying customers first, labelled early access. The general feeling about AI images at the time was that they were somehow both dangerous and a bit rubbish.

What was actually going on is simpler than the row made it sound. An image model does one job: you type a sentence and it gives you a picture. The difficult part was never the drawing. It was the judgement. Ask for “a nurse” and the model should probably show you a range of people. Ask for a specific historical scene and it should not. Google had taught it the first rule without teaching it when the first rule does not apply, and it took six months to unpick that.

Then watch what the next two Augusts did to it. Almost exactly a year later, on August 26, 2025, Google released a model everyone ended up calling Nano Banana. The change that mattered was not that the pictures got prettier. It was that you could now point at a photo you already had and talk to it. Take the sunglasses off. Put me in a suit. Make this the same shot but for Christmas. And the person still looked like themselves afterwards, which is the bit that had never worked before. In the two weeks after it launched, Google’s Gemini app picked up 23 million new users and people made about 500 million images with it. For a fortnight every feed was full of little figurine portraits. Now, another year on, the same model is a menu item inside Google Photos on both iPhone and Android, and it is in Search and NotebookLM too. Google says more than five billion images have been made with it, and the Gemini app is now turning out around 150 million a day. There is no announcement to react to any more. It is just there, next to crop and rotate.

Which is where it stops being a technology story and starts being a small business one. A headshot for the website, a product sat on a plain background, the same photo redone for the season, that used to be a person you booked and a bill you paid. It has quietly become ten minutes and a sentence. That change never had a launch day. It arrived as a button.

So the lesson is about shape, not about images. The thing that got pulled for being embarrassing is very often the thing sitting in your pocket two years later, doing a small job you used to pay for. The useful question was never whether it was ready. It was which quiet job it would have taken over by the time you next looked.

If you want to read more:

Yesterday’s breakthrough is today’s normal.

The newsletter

Get On This Day in AI every week.

One story from AI's recent past, what happened next, and what it means for running a small business. Five minutes, no hype.

Free 30 minute call

Tell us what's eating your week.

You leave with two or three ways to save time, whether we work together or not.