Monday, August 22, 2016

How my approach to perspective has changed

It's interesting to me to think of how my approach to perspective when painting scenery has changed so much over the past few years. When I first started making background art for games I largely ignored the use of consistent, fixed vanishing points; instead, I relied on simple atmospheric perspective (the act of painting things off in the distance closer to the colour of the atmosphere to show that there was more atmosphere between those objects and our eyes than things closer to us) to show distance, and loosely approximated architecture with what seemed about right to my eye.

The result was often quite wonky - mostly something I embraced quite happily. You can see below that the foreground objects and the background objects go off to very different vanishing points - something that doesn't matter too much here, as things in the world aren't always aligned neatly to perfect grids, but it also doesn't have a great deal of depth beyond atmospheric perspective, and therefore mostly exists on 4 layers - sky, buildings, fence and playable area.

This worked fine in Blackwell Epiphany; the perspective is pretty loose, but most of the scenes are simple and flat enough that it doesn't matter too much. It gave the added bonus of allowing me to do doors and furniture at any height and angle that I felt like, more or less, meaning I rarely had to worry about scaling. I could mostly just eyeball the angles and get it pretty close.

When I started working on Technobabylon after that project, though, things got changed up quite a bit. No longer was I able to freehand my lines and architecture; James would provide me with simple blank 3D models of the scenes he wanted, rather than rough sketches, and I painted over these to give him the exact shot he was after. Because these 3D models had perfectly rigid perspective, my wonky perspective stood out quite a lot, and I had to quickly adapt my methods to pay more attention to vanishing points.

This was the first scene I did over a 3D backdrop and you can see some of the issues - most notably the seam in the left hand pillar shoots up too high as it recedes from the viewer's eye, and the foreshortening on the black line I painted on the trunk of the car is terribly off. Foreshortening still gives me trouble, unless I measure it very carefully, and I only know how to measure it for very specific circumstances. Getting better at it is something I'm aiming for. The cityscape in the distance was also eyeballed, and the perspective is all over the place there. I wasn't worried about it being consistent at this point, just concerned with making it look good, and you can see that it doesn't match - I didn't even really think to try making it match the foreground perspective, let alone making it internally consistent.

Here's an example of how a scene looked in the original 3D model James gave me and how it ended up after I painted over it. The strong perspective and foreshortening really allowed us to show depth in the scenes in a way I'd never been able to get with my loose approach. It taught me that here was a tool for adding distance, beyond the easily implied, this gave a much more solid, definite feeling of depth.

We still had to cheat perspective a little bit, though, to make some interesting ideas happen. James wanted to have a tall, vertically scrolling room with racks of mechanical, humanoid skeletons in the background, and two flights of stairs to descend. In order to allow the character to walk on the walkways, we had to duplicate the same architecture without altering the angle - that is, one is copied and pasted above the other without any alteration in perspective. When viewed as a complete image, it's quite clear that the two don't go to the same vanishing point, and the stairs stand out quite a lot, as they change and suddenly meet a walkway that doesn't match them. however the scene is given an additional sense of depth by means of having several layers of objects that all scroll in parallax. I think this helped to sell the illusion a decent amount.

I found painting over 3D models like this quite challenging, but I also found it taught me new things.

Shardlight was a very different project - unlike the closed interiors that make up a lot of Epiphany and Technobabylon, this was a project in which we wanted to show vast areas. Not only did I have quite a few scrolling scenes to paint, but I also had to try and show a lot of scenery in one shot.

For these reasons, I largely chose 2 point perspective. Two point perspective helped me a lot with scrolling scenes; because the angles at which objects recede into the distance are less harsh, it allowed me to show angles and distances on these wide shots without having to keep everything on a foreground plane. The scene can wander off into the distance in multiple parts, which helps show a lot of the world in a single shot, too.

The bazaar scene is one example of a location where we wanted to make it feel like a big world, and where 2 point perspective helped. Rather than the very tight, confined street corners I'd painted in the past, such as the very first scene in Blackwell Epiphany, here I could show the perpendicular streets both going off into the distance and the various stalls that lined them without relying on having a flat area for one of the streets as I'd normally do in 1 point perspective. I like playing with how much perceived depth this can add to a scene, more than the more flat, convenient appearance of a street whose horizontal plane aligns with ours.

This alley scene is a nice example of how I learned to use this type of view to add depth. The overhanging section of the structure helped me to take a flat side of a building and bring it not only out, but over part of the scene, showing another plane that I normally wouldn't show in this sort of view. Doing this also works in 1 point perspective, but not quite as well, here you get the full illusion of a 3D image.

One fun part of Shardlight was this vision sequence, for which I ignored perspective grids completely. One thing that I find with using perspective that goes off to a point is that it feels very rigid, and it was fun to abandon that completely for these scenes and paint by eye once more. This is closer to my approach from before working on Blackwell Epiphany than most things I've painted for Wadjet Eye, and felt like a chance to allow my imagination to run free.

Using two points of perspective helped in this cathedral shot, too, one that was designed to begin at the top and scroll down, establishing the height of the structure. Francisco originally asked me to do this in 3 point perspective, but I felt it'd be stronger with just two; one at the top and one at the bottom, and I thought that doing 3 point perspective in such a large scene with such a complex structure would be a lot of work for me. I stuck to just 2 points, focusing on a fairly symmetrical look, broken up by the chaos of the destruction and repair of it. It was cool to try this sort of shot, and a bit intimidating - which means it was also educational.

Interestingly, the first few scenes I painted for Shardlight were done before I worked on Technobabylon, and you can see that they're painted in the looser, eyeballed style of my older work:

Things have changed once again now that I've been working on Unavowed. The change to 640x360 resolution meant that I was nervous - I was out of my comfort zone again, trying to learn how to paint things at twice the resolution that I'm accustomed to. I decided that trying to manage the more complex forms that two vanishing points require would be an extra burden that I didn't want, and so decided to stick with 1 point perspective once more - something aided by the fact that I'd mostly be drawing interiors and city streets.

One of the tricks that I tried to give more illusion of depth to this scene was to focus on very 3 dimensional forms in the structure of the ceiling supports, solid, thick shapes that help to establish an impression of distance and width. The fact that this area in my scenes is usually not where players will be wanting to examine things or be able to move means I can be more liberal with experiments like this, and I like the effect it gives here. I'd like to try this sort of thing more, and I'd like to learn to get foreshortening measured out more accurately so I don't try to eyeball it and get it wrong as I did here.

Something I've noticed in Unavowed - whether it's because we have more characters that can move around than ever before, or because of the higher resolution, or because I think about perspective more, or because I'm using so much 1 point perspective, or some combination of these things - is that characters go out of scale and proportion a lot. Doors end up looking enormous, characters end up looking tiny when they're placed closer to the foreground, and the sizes of things look inconsistent as a result.

The only way I can think to fix this offhand is to have either very, very careful character scaling, or to lift my horizon lines much higher, meaning the characters walking into the foreground can never get quite so close to the player's "eye" into the world. This worked well in Beneath a Steel Sky, for example, but my instincts and the various things I've read lead me to believe that painting with the horizon line closer to eye level, as I mostly do, is more immersive; providing that characters don't break scaling so much. It's another problem to solve and think about, and it's something that we've done for years, but now it looks worse to me that it did in the past.

One last thing to mention for now is that while most of my scenes with 1 point perspective have the point very close to the horizontal centre, sometimes I like to change it up by pushing it over to one side. I really like the effect that it gives, especially on a street corner like the scene shown about, where it's quite evocative to have the far distance implied just beyond the edge of the building there. I'd like to do this sort of thing more.

When I look back at how my approach to perspective has changed over the past few years (the screenshot above was painted about 4 years ago, and has no sense of consistent perspective at all) it's quite fun to see how tight perspective has grounded my scenery more in reality. I still have a lot to learn, and I've started to step back from the rather rigid perspective I use now a little and see what results I get after a couple of years of painting quite stiffly along grids. There are some avenues I want to investigate further, and no doubt plenty of things I have yet to learn.

Perspective is one of the fundamental elements to consider when creating 2D backdrops, and it's something I now like to pay close attention to after ignoring it for years. Perhaps in another year's time I'll have more thoughts and knowledge about it.

This post has the word 'perspective' in it 32 times, including the title and this footnote. I'm aware that's far too many. My apologies.


AJ said...

Very interesting read. Thanks for sharing it Ben. It's nice to know that we share similar methods in drawing backgrounds. For buildings, spaceships exteriors and interiors, I always paint over a 3D model. I only draw from scratch when it's an outdoor scene or nature sights like forests and so on.

Ben304 said...

I think this makes a lot of sense in drawing a corridor, or any complex, man made structure in which consistent, measured geometry is key. The solidity that using a 3D base provides is difficult to match. I rarely model 3D things myself, though will do it for certain things in which the foreshortening isn't immediately obvious to me. Having a reference from which to judge foreshortening is probably even more useful than having a reference for perspective, as perspective is much easier to measure.

polecat said...

Really interesting to read about how you adapt your approach. Loved your work on Shardlight in particular - you get so much atmosphere going.

Ben304 said...

Thank you! Glad you enjoyed the post, and thanks for the kind words. :)

Sankar said...

While researching for my own P&C Adventure Game, I noticed that older Sierra games almost always used 1-point perspective, and usually 2 camera angles for the whole game. LucasArts on the other hand, usually deployed a 2 point perspective setup and crazier camera angles.
But, on the other hand, Sierra backgrounds were way more detailed and usually with tighter lines, while lucas relied more on flat colors and cartoony lines.

So, I think your post captures this perfectly! The struggles you face with the ratio between construction/drawing and rendering.

On another note, I'm pretty surprised to find that you've struggled going higher res. I always picture your backgrounds being drawn in larger resolutions and shrunk down to fit the game resolution. Specially in Technobabylon.

Thanks for sharing and I hope to see more posts from you soon. Big fan, man!

Ben304 said...

Yeah, LucasArts often feels more diverse to me, just because of the range of styles their artists used over a shorter period of time. The likes of Peter Chan really knew how to mess with perspective and get away with it. Of course, you can see examples of this in Sierra games also, but not to the same extent.

I did resize a long time ago, but Technobabylon was definitely done at game resolution! I've been drawing at pure 320x200 for quite some years now, so the changeup was definitely something to get used to.

Thanks for the comment! :)

Anonymous said...

I'm glad you left this blog up, providing nice insight into the creative and developmental process of your artworks. Thanks.

Ben304 said...

You're most welcome! Thank you for the comment. :)