Anybody who has spent any time trying to render the same character in multiple shots over time with AI Art has run into the issue of consistency. It is hard to get your characters looking the same even twice in a row with the same prompt, much less with different prompts and different scenarios. There is one rather interesting way around this limitation in short bursts, however -make the AI generate several images at once. When it does that it can give you *exactly* the same characters, scenes, etc in several panels with differences between them. You still have the same word limit, so you can’t quite make scenes as complex as with four separate renders. But since you only have to describe the things the images have in common *once* its not as bad as it sounds. In this article I will explore some of the things you can do with this technique.

---

Variations on a Scene

Here is one example of a render I did for my “Solasta: Dimensional Discoveries” article. For this shot the conceit is that we have the exact same Arcane Library existing in several different dimensions – each one serving a different purpose and being used slightly differently. Getting four images looking different but representing the same exact scene in four renders would be an exercise in frustration. But by doing this, we can just repeat a few times until we get an entire set we like:

“fantasy painting;generate four square images of the same (underground marble arcane library): (first is clean and orderly with blue light);(second is collapsed and burning with hellish red light);(third is strewn with magic treasure and green haze);(fourth is dark with a huge ghostly translucent dragon)“

A variation here would be doing the same scene in four seasons:

“generate four images: first in summer,second in fall,third in winter, fourth in spring;jeff easley fantasy painting;ancient ruins of elven roman arcane academy nestled in hillside in pine forested countryside”

---

Reaction Shots

Here is another example from this same article. For this shot the idea was to get the character’s reaction changing from excited anticipation to concern as she reads what the parchment says. So I tried a few different variations of the prompt below, with the first picture (the one I used in the article) using the exact prompt shown.

fantasy painting;generate two images of (short stocky halfling woman,wide face) reads crumpled dirty parchment;(halfling woman,wide face,stocky build,age 35,short red hair tied back,green eyes,short,dark green strapless tunic,dark green pants,dark green cloak,leather boots);large bookshelf;in medieval stone drawing room: in the first she is excited, in the second she is worried“

---

Fight Scenes

You can use this technique to generate panelled sequence fight scenes. I haven’t experimented much to see how different the individual panels can get, but the simple setup below shows a glimpse of what it can do:

“action scene;fantasy painting;generate four images of: (rogue woman,dagger) fighting orc;(italian rogue woman,age 30,short black french bob hair,brown eyes,black leather armor,black cloak): in first image she charges, in second image she dodges, in third image she tackles, in fourth image she leaps“

Or we can get more sequential:

four images of rogue woman shooting bow at lizardman: first she aims the bow,second she fires the arrow,third the arrow flies,fourth the arrow hits the lizardman;fantasy painting; (italian rogue woman,age 30,short black french bob hair,brown eyes,black leather armor);(hulking black lizardman,sharp bony growths);in a dark dungeon”

---

Storyboarding

We can even start to tell a story – comic strip style. Its still inconsistent and you have to do a lot of takes to get exactly what you want. But check these out these variations on the prompt below…

“generate four images as (italian rogue woman) interacts with (french paladin man):in the first he gives her bad news,in the second she gives him a sour look, in the third she smile at him, in the fourth he hugs her;(italian rogue woman,age 30,short black french bob hair,brown eyes,black leather armor);(french paladin man,age 50,neat thin beard,friendly,haggard,brown hair,blue eyes,strong,paladin armor,blue cape);fantasy painting;in medieval tavern”

---