MJ's take on Daleks, both using the Niji and v5.1 models:
Fairly accurate.
Fairly accurate.
They get some details correct, and I can easily see the influence of Old Style Dalek designs (as well as Bronze Dalek), but it's the subtle details -- it looks like what an paleontologist would think a dalek would look like after seeing its bones, if it were an animal, for example.MJ's take on Daleks, both using the Niji and v5.1 models:
Fairly accurate.
No, I'm from the US, from the general vicinity of Seattle, and I'm of Spanish and Scottish extraction. Some of my best buddies from work are Polish, lol. Lots of Norwegian, Greek, and Polish dudes in the US Merchant Marine. I don't think any of 'em knew about Beksiński, though. I've been an H.R. Giger fan for a long time, and Beksiński's name often comes up in discussions of Giger due to some of the similarities in their bleak, macabre, organic, intricate styles.@*THASF* You are the first person that knows about Zdzisław Beksiński ... are you Polish???
If only...
Probably. You don't kill several hundred thousand people without repercussion, after all.Tell me one thing out of curiosity...is it true that Tito is hated more in Croatia and Slovenia despite he was Slovene and Croat?
Midjourney 5.2's take on that looks a little something like this:
I've tried doing yandere characters in MJ Niji and had them, you know, just adopt the infamous Yuno pose automatically, without specifically including that in the prompt. The prompt for this one included the phrase "an expression of sadistic glee", and that was enough to get Yuno poses, lol.Okay, here's a prompt that may baffle the AI, and I'm honestly wondering if they'd get the "pose" (you'll know what pose in a sec). right.
Yandere. Anime. Catgirl.
Let the gates to Tartarus open, and let the nightmares flow!
...For some reason, now knowing this and how AI operate behind the scenes? Yeah, I'm scared.I've tried doing yandere characters in MJ Niji and had them, you know, just adopt the infamous Yuno pose automatically, without specifically including that in the prompt. The prompt for this one included the phrase "an expression of sadistic glee", and that was enough to get Yuno poses, lol.
The Niji model (for cartoon/manga characters) is better at it because it actually knows what you're talking about (that is, it has more occurrences of that specific pose in the training data set). But yes, it can definitely 100% do that....For some reason, now knowing this and how AI operate behind the scenes? Yeah, I'm scared.
Does it do it with more realistic characters? E.g. Hermione Granger or Natasha Romanov Yandere?
How do you get such good space ships? Mine often turn out all lumpy and stuff.Alright, after taking some time unsubscribed from MJ I recently returned after v5.2 with the new features was released.
Mixed bag, even though some great improvements have been made.
Regarding niche franchises like Battletech, the algorythm does way better work when using the niji (anime) subset. The reason here is rather simple: you don't have all that many useful Battletech 'mech images only to scrape, but you have tons and tons of mecha from literally countless anime series and mangas. Ironically, niji also produces (to me, at least) a lot more visually pleasing space ships than the standard algorythm.
The zoom and pan features finally allow for some panoramic shots.
Example:
Base Image
Zoomed out (1.5 zoom)
Panned
Where I feel it's gotten worse is putting different concepts together. Whereas I was able to conjure up a cyberpunk street samurai in v4 or early v5 it now really concentrates on "samurai" and "street" rather than the underlying concept.