3 ms·
The article starts off with me expecting the twist to be that the image was generated with a single text prompt. That would have been neat, and in line with the
by tshaddox 4y ago
The article starts off with me expecting the twist to be that the image was generated with a single text prompt. That would have been neat, and in line with the other recent sensational coverage of how these new models are BLOWING PEOPLE’S MINDS. But it’s actually a walkthrough where the author goes through a much more tedious process that I could ever imagine going through to get the level of quality of that final result.
- SamBam 4y agoLike my sister comment, I had the exact opposite reaction. Dall-E is very cool, no doubt, but the idea that I could actually work alongside the AI to produce something that is in my head was much more eye-opening.
- FemmeAndroid 4y agoHuh. I had almost the exact opposite opinion. As someone who likes to make art, this example leads me to believe that this kind of process can allow me to quickly create images that are far closer to what I have in my head than most of the DallE style examples I’ve seen recently.
- vidarh 4y agoMy reaction too. I've only superficially played with some of the AI image generators, and while it's easy to get something which looks interesting or even good, getting something which matches a specific idea precisely seemed hard. This walkthrough shows a clear technique which seems endlessly adaptable to using it as a tool to get to a specific goal.
- barrkel 4y agoThis article demonstrates control via iterative techniques, which is more flexible than trying to encode everything in a sentence. The input image acts as state carrying forward much more information than a sentence from one iteration to the next.
- boppo1 4y agoI'm very skilled at illustration and fine painting. Compared to what it takes to "manually" make such an image, his process is about as trivial as a single text prompt.