Don’t ask me if it was too many YouTube ads about AI workflows or because my brother showed me something on his phone that made we think this was a good idea… I figured, during my next project, I’ll upload images to Google Lens via the Chrome app and use that as a feedback loop to see if things are going in the right direction.
Had a rocky start, it identified correctly that I held a 3D printed part in my hands. Fair enough, not a mind reader… after a while we had something that somehow looked like a Gothic or Victorian mirror. I squinted, and I could see it too. Eventually, I had the whole piece assembled and we had a Gothic clock.
However, I was not trying to build a Gothic clock, so I did some dry brushing and with more details coming out, we arrived at the conclusion that it’s actually more Empire style. My inner 90s Valley Girl just went “as if”… added some color to it once the .2mm nozzle inlays were done printing and we finally arrived at Belle Epoque via Napoleon III style.
Great, so far I learned nothing new… but maybe if AI can’t help “along the way,” maybe we can work out the last detailing, adding some highlights here or there, image composition, maybe even a few key points to include in the model description when I upload it.
Uploaded the final image and got… “Yup, that’s an antique clock.” My first reaction, “Oh thank you, I’ll take that as a compliment”. So I added the old 3D printing prompt “Why do I see layer lines” and now this just made my day…
…Google Lens starts channeling its inner Obi-Wan Kenobi and tells me “these are not the droids you’re looking for”… enjoy:
…yeah, I’d like to not share any further discussions we had after that. I’ll leave it at “these are lay lines not layer lines” and I most likely photographed an actual antique. I said that probably makes very little sense and I was told that lots of people on MakerWorld use renders or pictures of the actual item they work off of.
I asked a few people with more AI prompting experience and they told me that you can even use ChatGPT for write prompts for Google Lens. There’s a way to get it to accept first that you did design, print, post-process, and photographed it and then you go from there course-correcting along the way.
Honestly, by the time I’m done doing all that I can probably have my next model doing its first test print. So yeah, that was that little story. Did it waste a lot of my time… well, I used idle time while stuff was printing for all this, so not really. Did it help at all… nope, not in the slightest. It did have some entertainment value but more like a bad movie you’ll just watch only once.
If you’d like to have a look at the layer lines, the model is here. I mean, if you’re way better at this AI assisted workflow and think I shouldn’t consider it a failed experiment and call it quits… do me a favor and throw any of these images in there and show me a prompt that leads to an answer that would actually be useful feedback rather than AI cheerleading and congratulating you “calling out” (its words, not my) the mistakes from its responses.
