We’ve come a long way from the likes of Alexa and Google Assistant!

AI has been a hot topic over the past couple of years. From the creation of ChatGPT to the rise in voice cloning and music generation apps, more and more AI software is continuing to pop up.

AI learn from a process called ‘machine learning,’ where an AI is given encyclopaedias of information (or large data sets) to help them better understand a subject, allowing them to answer our questions and/or complete tasks based on the information they have access to.

Imagine you learned all the information out there about a piano. You could know everything from how to put one together to every brand which has used one, ever piece of music, artist, style and arrangement. You would likely be able to answer most questions related to the subject and create countless pieces on existing literature and styles. The challenge for us using AI is understanding how we ask it a question, or how we instruct it to create something to meet our desired outcome.

So below I created a test using Canva’s AI Image Generator to create a snowy Japanese village scene in the style of the famous Japanese artist, Utagawa Hiroshige. Using the produced image, I’ve placed one of Hiroshige’s works below as a comparison…

I submitted the following text to image description into Canva:

Painting of a snowy setting in a Japanese village, with traditional temples and trees interweaving the buildings. River in foreground. Include small figures of people in the distance. All drawn in the style of Utagawa Hiroshige.

As you can see, the AI was able to generate an image based on the style of Hiroshige, however there are some clear differences not only from the style, but also from my instructions.

The AI image has created an almost cartoon-like style with more use of vibrant colours than Hiroshige’s, there is only one figure on the foreground as opposed to more than one in my description, and it has given me a mountain with a straight-cut dark blue horizon above.

This example is merely to illustrate that using AI to get your desired outcome isn’t as easy as it looks.

It is likely that I would need to change my wording, look at the pre-designed art styles and other filters to get the outcome I would want to see.

My recommendation is to have a play and see what you get, as I find that this is the most productive and fun way to learn.

Below are the other designs I generated using the same text with different art styles. Feel free to browse and have a go yourself! – Jason