I asked ChatGPT to draw a bird a year ago, then did the same thing today. You won’t believe the difference!
If you ever needed a visual representation of how far ChatGPT’s AI image generation has come in just a year, then I don’t think there’s a clearer example than the two images I’ve generated and shown here, both from the same basic prompt, but created almost exactly a year apart.
The prompt I used was simply, “Draw a bird”.
As you can see, a year ago, ChatGPT was still creating images out of simple geometric shapes like triangles and circles:
Now, a year later, when I type in the same prompt, I get this:
The difference is like night and day, and it’s really hard to think that just a year separates these two images.
Free as a bird
One thing that ChatGPT is very good at is keeping a record of all your chats. Recently, I was upgraded to an Enterprise account, and in the process of switching, ChatGPT converted all my existing ChatGPT conversations to the new account, which is why I noticed one of my first interactions with ChatGPT from almost exactly a year ago, when I started my job on TechRadar writing about AI.
One of the first things I asked ChatGPT to do in 2024 was to draw the bird, just to see what it could do.
Of course, this geometric image isn’t representative of the state of AI image generation a year ago, by a long way. At the time, Midjourney and others were quite capable of producing photo-realistic AI images, but this was all my free version of ChatGPT could manage.
If I’d had a ChatGPT Plus account, things would have been different. A year ago today, I could have asked ChatGPT Plus to use its built-in DALL-E 3, a powerful AI image generator, to produce a photo-realistic image of a bird. It wouldn’t have looked as good as the one ChatGPT can generate today, but it would have looked a lot better.
Access to DALL-E 3 was added to all ChatGPT Plus accounts in October 2023, but at the time, it still wasn’t integrated into the free version of ChatGPT in the same way.
It wasn’t until August 2024 that free ChatGPT users could start generating images in DALL-E, and even then, you were limited to just two image requests a day. When you reached your image limit, you were back to generating images that look like the one shown here.
On trend
It seems almost unimaginable these days that ChatGPT ever produced such basic-looking images. Ever since ChatGPT has upped its image generation game with a new native image generation engine, back in March this year, we’ve come to take superb AI image generation from the likes of ChatGPT and Google Gemini for granted.
In fact, it’s rare that a week goes by without another ChatGPT-inspired image fad trending on social media. It started with Studio Ghibli-inspired images and went on to include the action figure trend and others.
Thinking about how far we’ve come in just a year is quite something, and it’s got me thinking about where we’ll be in another year’s time.
In 2026, will anything less than a fully realistic image with automatic video generation, including sound, based on the image be considered an abject failure? Quite possibly, or maybe we’ll all be wearing AI glasses by then, and we’ll be able to step inside our images, as if they were mini-worlds, and take a look at them from different angles.
Whatever is going to happen, I can’t wait to find out.
You might also like
link
