Log #11: Frustrations
Progress on the info product’s content has been frustratingly slow. When Thursday came around (having made basically no progress), I decided to write sections as individual markdown files rather than as part of a massive wall of text in Notion. That little change altered the trajectory quite a bit. Maybe the smaller scope helped with focus. Maybe just changing the interface I was looking at was a nice change. I’m going to continue like this for a while. I still don’t know what tool I’m ultimately going to use to format and produce the PDF. Probably Pages at this point.
On the cover art side, I upgraded to ChatGPT Plus to try out the new 4o image generation and see what I could get out of it.
The images are much better than before. But driving the AI in the direction you want is still difficult. And the AI often generates basically new images when you try tweaking a small part of the picture – I’d gotten super close to exactly what I asked for, but then the following image went off the rails away from the goal.
The cooldown and image limits were frustrating. I get it - but it takes you right out of it. And random failures didn’t help, especially when asking for changes with a mask. If it fails, you have to redraw the mask and add the instructions again, if you don’t, the AI doesn’t use the mask from before.
Uploading images to ChatGPT didn’t always work either. Too many times I uploaded an image only for it to appear blank to the AI, wasting an image generation credit. I still don’t know what causes uploaded images to not work.
When you do finally get something from the it, a majority of images are easily identifiable as AI-generated, particularly ones that aren’t super stylized or cartoony. AI is getting close - and it’s already good enough for some use cases (like scamming your Nan), but I’m not sure if it’s good enough for what I want.
Image generation is fun and addicting. And I’ve found a few styles that might work for the front cover, but the ‘standard’ for this type of info product is actually just a photo. I don’t have one I’d be willing to use.
I have mixed feelings about AI images (that is, AI image generation as a whole). Like, is it dishonest to use AI generated images? Does it cheapen the artifact you’re creating?
I may try to generate an image and then have an actual artist create it. But then I don’t know what that would cost. For now, I’m going to continue generating images and try finding something that really resonates.
The chair? It still hasn’t arrived. I asked for an update and they said “We will try to deliver the order by coming week.” Again. Quite annoyed.
What’s next
- Focus on the info product’s content.