I'm still heavily exploring these new tools from an artist perspective. I never managed to get a run on Midjourney, but between DALL-E and SD there are quite a few differences. Broadly speaking, DALL-E seems to better get a hang on photographic results and interpreting "what I meant". With stable diffusion it's a lot of fiddling and putting manual emphasis on certain keywords until just getting it right.
I went about trying to utilize stable diffusion for an imaginary concept project (concept for characters of a remake of TMNT, heh). Process was similar to how I'd do it with another artist more than if I drew it alone. It was back and forth, from rough outlines and then honing into details. Inpainting and img2img helped A TON and I hope I'd get dreambooth running soon as well since that will be a game-changer in the combination of things.
Between exploration phase, detailing, alternatives, and manual painting and over painting, I'd say PER OUTPUT final image I created in the region of a thousand or so interim images. Process overall did take a lot of time but not as much as completely manual and I didn't feel like I had as much control as manual of course, but I did feel ultimately bold enough that I thought I had creative control. With dreambooth I expect it to close the gap.
Overall, I was extremely pleased with the experiment and I'll continue exploring it, even though I'm not doing artwork professionally anymore. And so far no, it's not going to replace artists. It's another tool removing labour, but adds time on direction needed. Ultimately it'll be another brush in the toolbox.
Overall, pricing will need to be adjusted over time as well. I set out on an experiment the other day that you can see here: https://twitter.com/Keyframe/status/1574338738808934400
I went about trying to utilize stable diffusion for an imaginary concept project (concept for characters of a remake of TMNT, heh). Process was similar to how I'd do it with another artist more than if I drew it alone. It was back and forth, from rough outlines and then honing into details. Inpainting and img2img helped A TON and I hope I'd get dreambooth running soon as well since that will be a game-changer in the combination of things.
Between exploration phase, detailing, alternatives, and manual painting and over painting, I'd say PER OUTPUT final image I created in the region of a thousand or so interim images. Process overall did take a lot of time but not as much as completely manual and I didn't feel like I had as much control as manual of course, but I did feel ultimately bold enough that I thought I had creative control. With dreambooth I expect it to close the gap.
Overall, I was extremely pleased with the experiment and I'll continue exploring it, even though I'm not doing artwork professionally anymore. And so far no, it's not going to replace artists. It's another tool removing labour, but adds time on direction needed. Ultimately it'll be another brush in the toolbox.