HUMAN
AI
HUMAN vs AI
So, now it’s a few months later I decided to try another human vs AI comparison using one of my traditionally created photomontage and manipulated images for the human created image vs trying to create something very similar using AI. You can see the results side-by-side in the comparison above. I deliberately chose something a little less complicated this time and also thought that there would be plenty of reference for AI to make sense of the prompt.
It is also worth noting, as will all the AI generated images on this site, I present the AI image as it was output with no post processing or further editing to change things or remove anomalies. I have chosen to keep true to the AI original output from a prompt.
Of course, I have to caveat this all by saying that my photomontage and digital editing skills are probably far superior to my prompting skills, so the comparison will be bias towards the experience I have with non AI digital art. My original digital image (see above on the left) was made using ‘traditional’ methods of photomontage techniques, layers, blending and editing and took a number of hours to manipulate. As with the previous time, the AI version (on the right) took very much less time to create, but still many prompt rewrites and edits as well as discarded generations before I settled on the final one you can see above.
However, even though this was a much simpler image, I still found that AI was unable to achieve to a very accurate and similar look and composition. The play on the scale of objects seemed to be the most tricky aspect with many attempts needed before achieving anything close. A more experienced prompt engineer with a different or more defined AI learning model might well be able to achieve some better results more quickly.
Note that I didn’t use a reference image – the same as for my previous attempt. Including a reference image may have enabled AI to understand the scales and context better for the output, but equally could have restricted the variations, style and composition of the final output.
I have included larger versions of both images below so you can see the details a bit better and compare what went well and what didn’t go so well.
It’s a Big Wide World – human creation
View my AI art galleries.
