OpenAI's DALL-E 2 creates variations of my hand sketches.
So I'm testing DALL-E by uploading my own drawings, and seeing what it does and what it comes up with.
We click on upload an image.
I'm gonna grab a drawing of a building and I'm gonna click on generate variations.
Right. Those look pretty good. Pretty bad, actually. They're like not, not great. I mean, They look similar from afar, but they have nothing, nothing to do.
Let's try with this one. So this is a sketch, so let's try this one. This is a sketch of a notebook or a sketchbook, and I just click on generate variations. So we see what DALL-E comes up with.
So these are our results. You can see, that this is the original drawing that I have sketchbook, scanned and edited, and these are other variations. So it also seems like something that opened. It's understanding it's something like something that can be opened. Huh and, huh, it's funny. It doesn't really quite get it, but it knows that it's something that's open.
Okay. This is really interesting. I think these variations.
So these images actually look really similar among them. If you looked up them from afar, but if you look closely. It didn't quite get the context of what those things are.
So that was it for DALL-E. We just saw another set of like generations.
Please like the video to support the work I'm doing here. Make sure to subscribe, to get notified when I upload new videos or go live next and join the Discord community.
I'm looking forward to hearing what you think.
Thanks a lot for watching and I'll see you next time.