I tried to get a self-portrait made by Craiyon. I wasn’t quite sure how good their engine would be versus other “creative” prompts from Stable Diffusion (which I’m usually impressed by). I have to confess I have mixed feelings about my results: while it does capture the literal essence of my prompt, the panels are not different enough from each other and has apparent limitations. First, the face rendering is still deficient, not sure if this is because of the lower resource limitation based on this being free or if it’s a technical challenge for the model. I also may be picking on technicalities but it feels like it went for book stacks rather than just papers — but again, a book might be a stack of papers. I’m surprised how similar all panels seem, maybe it doesn’t try enough exploration/interpretations from the prompt and rather try to provide minor variations of the same interpretation. Surprisingly, they are all women — not the kind of bias I expected. But overall, it was a fun experiment on where AI is going next.
I do like how they are actually trying to drink the cup of coffee though!