So you upload the screenshot into MidJourney to give it a base right? Have you ever tried using img2img with Stable Diffusion?
I love it! This brought me back to my high school history class decades ago. I can hear the teachers voice going on about the Visigoths and Ostrogoths. That was probably my favorite class. Core memory unlocked!
I'm interested in hearing more about your workflow.
Yeah, I've seen some good generations from others, as well as some bad ones like mine here. I think it depends a lot on the prompt and I just put in "a cow eating an old man" with nothing else.
Yeah, but currently only via their API. The weights should be coming out sometime later. Also I've heard that SD3 is still undergoing training, so this current API version will probably be a bit different than the released weights.
he even took a knife out and stabbed it
We humans aren't going down without a fight, the bovines need to earn their revenge!
I wanted to see what SD3 would do with that prompt...
Great hand! Dalle I assume?
That must be an expensive purchase! I bet she's going to be paying back that debt for a while .
I added an extra note to the post ๐
Although I notice we already got a nice non-gun entry a few hours before you even wrote this.
Hmm, perhaps I should remove my two bonus examples? I was just trying to show what MidJourney and Dalle can create based on my own prompt of trying to recreate the scene I mentioned.
Let's see your version of payback ๐
Wow this is a dark one!
I took a look at your workflow and narrowly escaped an aneurysm ๐ I've had some workflows turn into something like this as well.
I haven't tested these yet, but was recently looking for regional prompting solutions in comfy to match what's possible in a1111/forge and found these:
I'm curious if any of these could simplify things.