Fixing the Workflow
Almost as much as writing an actual post, I work on creating an illustration for my articles. More than two years ago, after the Creative Commons images I used were flagged as copyrighted images and I was being threatened by copyright trolls, I decided to use generative AI images. First, I used very basic images from Craiyon and as new tools were released, I migrated over, even switching to a paid version of Midjourney.
Recently, I moved back to DALL·E 3 but was faced with a hiccup. The images are created in .webp format, which isn’t supported by Medium. I’d need to use Preview or some other program to export the image into a format approved by Medium, in most cases, .PNG. However, this is where making tweaks to ChatGPT can help.
“Please convert the image to .PNG. Thank you.”
ChatGPT responded first something like “Hey, it’s already .PNG” Um, no it ain’t.
Asking it again, I helped the system understand its error and it then obliged. A .PNG! Great… could I make the workflow even better?
In today’s image, I asked it to create a .PNG from the outset. It didn’t work. Asking it again, it was able to correct itself. I netted a gain of one turn in our conversation.
While AI is almost magical, most of the effort to get an output is in setting up the system and then building a workflow. We should always reflect back after using these systems… how could this be done more easily?