I was so excited last week, when I asked ChatGPT to edit an image for me, and it did!
But now, I’m beginning to think it was a fluke.
(Pretty much like being able to co-write The Remote Worker’s Guide to Time Management with it, I haven’t had similar results yet with any of my other attempts at co-writing.)
Come on the journey with me, with its highs and lows…
Music to indicate time travel or flashback goes here...
I needed a feature image for the post Memoir: Where Shakespeare Meets Remote. It was a long post, and by the time I’d finished, my brain was unable to come up with some kind of drawing to go with it. Knowing that it usually takes me ages to find the right stock image, I turned to ChatGPT’s DALL-EE.
First I tried to feed it a prompt that Claude had generated after reading my post. It refused to create the image. (I will come back to this in another post.)
Then I asked:
Can you draw a cartoon of a person thinking about events with other people and those events going into the writing of a book?
It created:
I quite liked the image, but wanted to customise it to my gender, so I clicked on the image, which opened up an Editor. I selected the whole image and asked:
Can you do something similar, but with a young woman writing, and not so many people in her dreams.
Which resulted in:
The dream was still very busy, but, hey, it worked for me!
I was quite excited about this (yes, I know, I should mix with real people more often…) and I thought: yes! It’s finally here, I’ll now be able to edit AI images without pulling my hair out!
(For more on pulling my hair out, check out When’s a Spoon not a Spoon.)
A few days later…
I posted my Wicked Wednesday post on LinkedIn. This is an experiment in posting ChatGPT work-based humour. The bot has created both text and images. I posted this:
"I put the 'pro' in procrastinating at my home office."
My friend Chris, commented:
”Ahhh….you’ve been peaking into my home office, haven’t you! Except cat instead of dog.”
Ok, I thought. Let’s recreate Chris’s space more faithfully…
So I went into the editor with the image above, selected the dog and instructed:
Turn this dog into a cat.
Followed by the typical elaborated GPT-4 response of:
I've transformed the scene to feature a cat in your whimsical home office paradise, emphasizing the cozy yet distracting essence of working remotely with a feline companion. This image humorously illustrates the playful challenge of maintaining focus amidst the comforts and distractions at home.
My first reaction was, where’s the cat?!
But now, looking at it with fresh eyes, I can see what happened.
The bot rephrased “Turn this dog into a cat” into “transform the scene to feature a cat”, which, by removing the dog, is exactly what happened!
Anyway, I continued on my quest to replace the original dog with a cat.
You removed the dog, but didn't replace it with a cat...
To which the reply was:
Nooooooooooo!
That’s not what I wanted!
(Though I quite like “Procrastation”…)
I went back to the editor, selected the dog in the original and said,
Draw a cat in the selection, playing with the pad
Ok, not what I was hoping for, but finally, we got a cat in the right place, even if it’s Schrödinger’s.
If you would like to discover more about using and not using genAI for creative endeavours through my trials and tribulations, subscribe below.