Developers from Cardboard Sword talked about the tools and techniques they’ve used to create the incredible ‘stealthvania’.
Unreal Engine 4 is not the kind of technology you usually associate with 2d pixel art platformers. However, it is this engine that Olly Bennett and Cardboard Sword chose for their next game The Siege and the Sandfox. It’s a lovely old-school product, which tries to combine rich 2d-visuals and stealth mechanics. In our long interview, we discuss the production of this project and various ways the team benefited from the use of UE4.
Olly Bennett: Aidan and I set up Cardboard Sword in 2013 after we finished working as producers on Sony’s LittleBigPlanet PS Vita. We had both spent around seven years in AAA by that point, and so wanted to try the freedom of an indie approach to development. We spent the first two years raising capital through consultancy work, all while concepting a number of game ideas. We were waiting for the right time and right idea before we started pre-production on one of them. That time came when a few of our ex-colleagues from other AAA companies also wanted to make indie games, and for very similar reasons. Ed, Keith, Chris and Oscar, as well as Jet Stone Studios, joined the team, and in late 2015 we settled on The Siege and the Sandfox.
Taking the indie leap was equal parts exciting and terrifying. It’s a big risk for us all but the potential rewards make it all worth it. We’re doing our own game in our own way, and if we break-even or sell well, then we’ll be able to keep making games that we’re passionate about in the way that we want to make them. That’s always been our team’s dream.
Aidan Howe: There seems to be a kind of conventional wisdom that Unreal is bad for 2D. To us it looks like very few people have taken up the challenge, so there’s little evidence either way. Our only gripe was that Paper2D is currently lacking a few labour-saving features and a bit of polish, but we’ve largely managed to extend the tools ourselves to solve that. Outside Paper2D the Unreal toolset is simply phenomenal.
With Blueprint we have been able to rapidly prototype AI without needing code time, so our engineers could concentrate on things like game camera and player controller. Those are also fair gripes with Unreal for 2D: the orthographic camera doesn’t play well with the lights without some gentle persuasion, and animating sprites has required quite a bit of custom code. Beyond that though, you have Blueprint, the material editor, and all kinds of very powerful tools that significantly shorten the distance between a standing start, and great visuals and gameplay.
The decision to go with Unreal was pretty much unanimous. The team is primarily experienced in 3D pipelines, and our designs for future titles plan to put it through its 3D paces, so part of the decision was thinking of future requirements. Really though our results thus far are just Unreal straight out of the box, with some custom character and camera code. Everything else is stock Unreal Engine 4 (lighting etc.) Game Maker was used for early prototyping, but we saw a lot of power in UE4 that we wanted to take advantage of, so transitioned pretty early on.
Chris Wilson: We started working with Ian Thomas of Talespinners on the writing from day one. He’s doing a great job creating our story and locations and the back-and-forth communication has made world building so much easier. With this narrative and plot overview, we have been able to create a white-boxed map of the whole game world. It’s basic but it gives us an idea of scale and overall layout, and forms the basis for iteration, which is also rapid.
From that we have a good idea of the general journey the player will make through the game, so we start to establish difficulty curve and unlock structure. We can then build each area in more detail and test it mechanically. The artists work collaboratively with me here to help design the visual layout of the space, such as adding important structures and themed objects that may dictate layout. Again, it’s iterative, but you build on top of the previous work until it makes sense and feels right. Once the gameplay elements are perfect, the artists move in to paint the area and make it look great. Good writing, good design, and good art all combine to make our game work.
Ed Duke-Cox: Well-crafted 2D pixel art has lots of charm; it’s timeless and so will always look great. We can draw inspiration from the games we played and loved as kids in the 80s and 90s, such as Chrono Trigger, the Metal Slug series, Final Fantasy, and others, mimicking these classics but with more modern techniques and in our own style.
The entire game obviously has to be constructed from modular tilesets, as this is a time-proven method of quickly and efficiently creating environmental spaces. We hope that by making effective and intelligent use of these tilesets, in conjunction with the layer and lighting systems available to us, we can imitate the impression of much larger, more bespoke hand-painted assets, but without the huge amount of time that would undoubtedly require.
To help communicate distinctions to the player, all enemies make heavy use of the same colour palette hues (same blue and/or red on most enemies) while friends all have heavier use of warmer colours like cream and brown. This consistent use of visual language permeates all aspects of the art throughout the game as well. Stronger saturation of colours used on characters and foreground environments compared to the muted backgrounds also helps to pull them forward and ‘pop’ them off the screen a little. This means players can use this visual language to tell at a glance what is a solid or interactive part of the gameplay space, and what is merely background dressing.
Light and darkness give us a huge advantage when painting backgrounds. By having everything recessing into darkness, and being able to focus detail into the better lit gameplay areas, we can avoid huge repetitive areas of tiling textures that plague original pixel games from yesteryear. Scenes actually become more interesting for having isolated hot spots of interest, that give the perception of detail, which isn’t necessarily always present.
One of the most important aspects of the animation pipeline was to attempt to capture a feeling of actual weight and momentum to all of the player’s actions (much like the Prince of Persia and Assassin’s Creed games). Everything from simple platforming to a Ninja Gaiden style wall-jump needs to ensure the player feels the physical impact of their movements. This allows the player to have a more grounded connection with the environment they are traversing. We rough out animations with silhouettes first, to get a rough idea of movement, flow, and ideal frame-count, before painting and detailing them.
Ed Duke-Cox: The environment art and characters are painted with an assumed front-on and slightly top-down light tint. This works well when we throw UE4 lights on top. We then add a light circle over light sources, such as sconce torches, that show the player the range the light is thrown.
We use triggers in these light circles to change how the player/enemy characters are rendered, dependent on their position relative to them. We shift the sprites’ colour values to represent whether they are inside the light’s influence or not. This helps to telegraph to the player whether or not they can be seen – i.e. hidden in the shadows or not.
We also add ambient light effects, such as moonbeams through windows or gaps in the environment to add more interesting light sources – like blown holes from siege engines or natural holes in cave sections, etc. – in addition to the environmental lights.
We tested normal mapping. The lion’s share of the fundamental pixel art had already been done by the time we moved over to Unreal; several environmental tile sheets and thousands of frames of character animation. No matter what clever solutions or quick fix software there may be available out there, the decision to employ normal mapping would require a significant amount of time and work to setup for use with all existing art assets.
The relative lack of resolution and screens pace available in pixel art games compared to modern games means normal mapping works most strongly on clean bold shapes with obvious angle changes in the geometry to show off the lighting. Siegefox generally speaking has the opposite of this. Lots of detailed and cluttered surfaces with micro detail such as crumbling brickwork or intricate engraved wall patterns. This meant that while detailed lighting could be achieved, the stronger the effect of the normal mapped lighting on the pixel art was, the more the ‘purity’ of the pixel art was lost. With normal mapping it wasn’t so obvious at first glance that it was a retro pixel art game, which is very important to us that we keep.
Aidan Howe: It’s always going to be very difficult for us to put a number on how long this game took to make, even once we are done. We had a lot of assets ready to go when the project was switched from Game Maker to Unreal, and work was divided between projects until we went full time on Siegefox early this year. If I were to make an educated guess I would say somewhere in the region of 15 man years, spread across 3 years, but mostly concentrated in the current and coming year.
That touches on what is most important. Some tasks are more time consuming than others, but that’s not something we can change. What we can do something about is ensuring enough time and the smooth flow of development without blocking work. Often production schedules are a bit unrealistic and solve this problem by putting pressure on developers to work harder and do more hours each day. We prefer to give as much time as possible and put the pressure on production to ensure that tasks are properly sequenced to avoid development being blocked or people burning out. Being independent developers gives us the luxury of time, so when planning I always assume the worst and give tasks twice as long as estimations. Our team are all experienced professionals, so if deadlines are not met my planning is at fault, not the work ethic or ability of our development team.
Though we’ve already mentioned it, tools like Unreal really are a major factor. Having access to a feature rich, stable, and regularly updated set of powerful tools, when placed in the right hands, results in a level of productivity unprecedented in our previous experience.
Olly Bennett: There are obviously lots of approaches to marketing for indies; some of them are necessary, such as the typical social media channels, but others can be hit or miss. There is certainly a lot of luck in who sees your content, where, and when, but you can improve your chances by sharing in as many places as possible and with networking.
Knowing what is unique about your game and how to express that concisely is important. Building from that, you need to know your audience and where they go for information. At those places updates must be regular, and scheduled as part of your workload. Squeezing bits in ad-hoc around development probably won’t work on its own.
Going to expos and conferences, and doing talks and networking will also help you to build contacts and locate opportunities. Try everything once and gather data as you do.
Our art is done to a very high standard, and stands out, which has brought in a lot of attention so far. One shouldn’t judge a book by its cover, but so many do – we’re confident we’ve got a good book too though! Hopefully, that will be reflected in our trailers, screenshots and demos. A good review or Let’s Play can sell thousands of copies alone, so we’re hoping to appeal to these content creators too. Also, as we release more and more content – including through sites such as yours – then we hope to find an interested group of fans that become vocal supporters, and that we can engage with throughout development.