Zeno's Ziggurat


RPG characters with AI image creation

I claim no ownership or copyright of these images whatsoever. You may download and use them for whatever purpose you wish.


AI Concepts: Perspectives

So in the course of illustrating Keira and Kord’s run through the BG Saga, I’ve learned a lot about perspectives and framing when using DALLE-3/Bing Copilot. So here’s a “concepts” page playing with framing and perspective.


Keria

So we’ll start with just a basic picture of Keira, no frills as our “control” baseline.

“half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt white hair,pale skin, purple eyes,black leather armor,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

This gives you a fairly typical “RPG Portrait” sort of shot. Now say we wanted a full body shot. You can try asking for that directly – and it may or may not listen. But if we are clever, its easier. Simply also mention something that can only be seen if you do that – like footwear!

full length picture;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt white hair,pale skin, purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

Alright. But the background is rather boring, so lets set things up a bit. Note here that this can immediately start affecting the camera view as well!

“full length picture;in the forest;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

Note that only one of these is now a full-length picture, and we’re getting more variation in looks. With pics like this you always have to set the stage first, then add the players.

full length picture; woman walking in the forest;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

Better, but still not full length. Now we start playing with explicitly directing the camera. Start with the obvious – pull back!

wide angle view of path through forest heading towards distant mountains;woman in the distance walking towards the camera;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

Note that even with all that it still doesn’t really want to pull back. Now lets play with it some more.

birds eye view of woman walking on path;path through forest heading towards distant mountains;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

Notice here how even with the high camera angle it really wants to zoom in on her. Now lets change the emphasis a little.

panoramic view of forest path,next to lake, heading towards distant mountains;half-elf woman walking on path;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

Side views are harder, and you have to fight for them sometimes.

distant view from the left side of half-elf woman walking on path;path is a winding path through a forest,with mountains in the distance;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

We get a bit of a side view, but not full length. Just adding “full length” to the description helps, but its still not consistent as either “full length” or “side view”.

distant full-length view from the left side of half-elf woman walking on path;path is a winding path through a forest,with mountains in the distance;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

So what if we change the action a bit to emphasize what her feet are doing?

distant view from the left side of half-elf woman walking on path;woman is idly kicking pebbles as she walks;path is a winding path through a forest,with mountains in the distance;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

Hah! That’s showing those pebbles!!! Lets be a bit more subtle.

distant view from the side of half-elf woman walking perpendicular to the camera;path is a winding path through a forest,with mountains in the distance;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

At least this is consistently sideways. Now lets narrow things down a bit and emphasize the “sideways” bits.

distant side view of half-elf woman walking perpendicular to the camera;full length shot;mountains in the distance;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

And what do you know? Three out of four is pretty good for this sort of thing. Now lets get fancier with a layered approach:

extreme long distance side view of half-elf woman walking perpendicular to the camera;full length shot;tall grass in the foreground;mountains in the distant background;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

Now say we want to push her back even further. How about we give the AI/camera something else to focus on?

extreme long distance view of tiny mouse hiding in the grass;tall grass in the foreground;mountains in the distant background;half-elf woman in the background walking perpendicular to the camera;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading

This technique can be used effectively in lots of scenes where you want it to pull back.

extreme long distance view of green parrot sitting on a pirate ship at sea;rigging in the foreground;half-elf woman far behind him in the background,looking out to sea;half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek;oil paint,drawing,dark fantasy,dark shading”

That one of the parrot in armor is pretty crazy, but by giving it something to focus on in the foreground we can really push our main character back and show more. Its even more effective if we cut down on the detail so it doesn’t get pulled to strongly towards the “main character”.

long distance view from behind of (dark-haired,bearded male pirate captain);captain is watching a battle below him;in the background on deck a (half-elf woman,french,short unkempt bob white hair,black leather armor,two daggers) is dodging and fighting pirates;oil paint,drawing,dark fantasy,dark shading”

We can push him to the side a bit as well:

long distance view from the side of (dark-haired,bearded male pirate captain);captain is on the left watching a battle below him;in the background on deck a (half-elf woman,french,short unkempt bob white hair,black leather armor,two daggers) is dodging and fighting pirates;oil paint,drawing,dark fantasy,dark shading

We can also play with getting it to show more of the scenery around a single character, by emphasizing that more:

long distance view from above of deck of pirate ship;rigging in the foreground;in the background a (half-elf woman,french,short unkempt bob white hair,black leather armor) is looking out to sea;oil paint,drawing,dark fantasy

Or this one:

view of small crab on beach;rocky beach in the foreground;pirate ship in the ocean far in the background;half-elf woman in the distance waving to the viewer; (half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek);oil paint,drawing,dark fantasy

Sometimes we can even pull off something like this:

long distance view of small crab on beach;rocky beach in the foreground;pirate ship in the ocean far in the background;half-elf woman in the distance waving to the viewer; (half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek);oil paint,drawing,dark fantasy

Or this:

wide angle panoramic view of rocky beach;sand in the foreground;pirate ship near the horizon;half-elf woman back in the distance,waving to the viewer; (half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek);oil paint,drawing,dark fantasy

You can also use the perspective to give a real sense of distance and scale:

extreme long distance view of rocky beach;sand in the foreground;a half-elf woman can be seen looking out to sea;(half-elf woman,age 22,french,thief,mischievous grin,scruffy,athletic,cocked head,short unkempt bob white hair,purple eyes,black leather armor,black buccaneer boots,small tattoo on her right cheek);oil paint,drawing,dark fantasy

If you don’t call too much attention to specific characters, you can do things like this:

top down view from extremely high in the air of rocky beach;very far below a fight can be seen between adventurers and pirates;oil paint,drawing,dark fantasy

Or even introduce rough characters back into the scene:

wide scenic view from high in the air of a rocky beach;in the foreground are jagged rocks;in the background a (half-elf woman,short white hair,leather armor,two daggers) is fighting a pirate captain;oil paint,drawing,dark fantasyher armor,two daggers) is fighting a pirate;oil paint,drawing,dark fantasy

It ain’t perfect, but you can certainly work at it to get some interesting views.

There are other views – like satellite view – which just don’t seem to work for me if I describe people at all. Bird’s eye view can work, but again, the more you detail the people in the scene the more it pulls in the view to show that.


8 responses to “AI Concepts: Perspectives”

  1. You seem to be able to work out the descriptions, to get what you want. Obviously I’ve had some success getting such things, but I get the impression you get acceptable results with a lot fewer attempts than I do!

  2. I find it’s a lot like working with children. Sometimes instead of making it do what you want, you figure out what it wants to do and why, and then decide how to make best use of that.

    But I’ve been working with AI since the 90s. And I’m a father of three. I think that way without thinking about it.

    It does help that it appears to understand photography and art terms like “foreground” and such. My daughter has been taking photography classes in school, so I just listen to how she describes things and try that. Your wife probably understands the same concepts.

    1. She does, and she has done a couple of AI works now. I don’t think her heart will ever be in it though. She has helped me with terms a few times.

      But it sounds like you’re saying, if we’d only had kids I’d be better with this?

      Given the tendencies you’ve seen with this AI art engine, would what you’ve learned transfer easily to a different program, or are they all completely different?

      1. Heh. Nah. Kids is just one way to get experience working with willful entities that don’t think like an adult. From the way my brother (childless) talks, experience in customer service is another. 😉

        The biggest difference I’ve found between Bing Copilot and other engines is that it is *very* *very* good (comparatively) at translating text into image details. There are engines out there that let you do fancier things – like dividing the image into sub-areas and tasking each separately. Or like training characters and referencing them. But it takes a lot of effort, skill, and API experience to get them to do that. Copilot/DALL-E just works. Take the exact same prompt and drop it into other engines (unless they also have a DALL-E interface) and they rarely do anywhere near as good a job at catching the details.

        So TLDR – Copilot isn’t necessarily as good at the “generating imagery” part as some tools; but it is one of the best at the “understanding what you are asking for” part.

      2. I will keep that in mind! I have an idea for a run, it will be a few months before I get to it, involving a couple characters from a favorite TV show. I’ve played a little with Meta’s Open Art as a “proof of concept” sort of thing. Just enough to convince I myself I can make it work. At least on a basic level. So maybe around Christmas time I’ll see what I can really do with it.

      3. I’d be very interested to see what you can do with that. I don’t think I’ve played with that one yet. I’ve heard “Grok” has some interesting capabilities as well, but I haven’t really gotten around to digging into it very hard.

      4. Meta’s let’s you input pictures for characters, so that’s its selling point to me.

        But I’ve got a couple others I’m eager to get to first!

      5. I’ve tried a few that let you start with a seed picture. Can’t remember the names offhand. But I suspect most of them have their strengths and weaknesses. Just having a “negative prompt” that works would be huge. E.g. “NO BEARD”. Or “NO POINTY EARS”. I’ve tried a couple sites that provide this, but I’ve yet to be convinced it actually does anything.

Leave a Reply to atcDaveCancel reply