Unusual Evolution – Developing Space Station Silicon Valley

Body swapping in games is a fantastic if somewhat underused mechanic. Actually, it should really have its own genre classification. The chance to gain the abilities of any foe, if only temporarily, or become an indigenous creature uniquely suited to a specific environment, has fantastic and almost limitless potential. There have been some good examples over the years, with Space Station Silicon Valley generally regarded as one of the best, receiving decent review scores in magazines and fondly recalled by N64 owners. Besides being a fine accomplishment in itself, SSSV also represents the difficult shift into 3D that developers experienced in the mid-1990s and, even more significantly, was an influence on Grand Theft Auto III and the entries that followed.

This was at a time when Rockstar North was called DMA, and was best known for creating Lemmings. Despite SSSV's later influence it was a humble beginning, with many of the team newcomers to game development. Programmer Grant Salvona recalls: "The SSSV team all started the same week in September 1995, at DMA in Dundee. I started the same day as [programmer] Obbe Vermeij and [lead programmer] Leslie Benzies. As none of us were familiar with Dundee we looked for accommodation together and ended up renting a house in Lochee - one of the rougher parts of town. We would get to work in Les' car around 9:00 and still be there at 21:00 most nights. Les and I would head to our hometowns on weekends, leaving poor Obbe to the horrors of the Dundee nightlife. The team consisted entirely of new starts - with Jamie Bryan a long-term DMA artist heading up the team."

Programmer Obbe Vermeij concurred with Salvona, also describing their work environment: "I mostly remember how basic everything was at DMA. For the first two years we didn't have a proper office and were working in a warehouse. Just before 17:30 somebody would reverse the DMA van into the warehouse/office so that the wheels would still be on in the morning - Dundee is not a nice place. I also remember the audio technicians building a soundproof room out of old mattresses."

Many of the new recruits, such as Vermeij and Benzies, would later work on the GTA games when DMA became Rockstar North. Benzies went on to head up the company, and is described by some as the "mastermind" behind GTAV. Vermeij meanwhile would become technical director on the series. As he reveals: "Most of the people that worked on SSSV ended up working on GTAIII and they took their philosophy with them. On SSSV we wanted a game world where the player was encouraged to try stuff to see what happens (for instance, in one of the first levels there is a sheep that eats flowers and craps them out again). We wanted a game world that wasn't just about reaching the end but was alive and interesting to explore. To be fair, with the limited time and team size we had, we fell short somewhat. For GTAIII we attempted the same and succeeded."

Some animals are "allergic" to water, taking damage when submerged. Tread carefully, or find something waterproof.

For those who have never played SSSV, the premise is that the eponymous space station, Silicon Valley, mysteriously disappears and then reappears a thousand years later. Launched as an experiment in robotic evolution, which has gone unchecked for a millennium, it's now populated by a diverse menagerie of wacky synthetic animals. The player, in the role of an autonomous "robotic control chip" belonging to a robot called Evo, must deactivate and possess these animals. Every animal has unique abilities, from rats which excrete explosives to purple gorillas which can swing on vines, and all must be used creatively to solve environmental puzzles and defeat further animals. A gun-toting desert fox, for example, is strong when fighting against a vulture, which in turn can fly to reach switches the fox cannot. Throughout the game there are interdependent puzzles requiring clever animal switching. In reaching this finished state, however, SSSV underwent many revisions.

SSSV started as part of a three-game publishing deal DMA had done with BGM Music – the other two games being the original Grand Theft Auto, which everyone knows about, and Tanktics. Salvona explains: "The original plan was that all three games would be released on PC, PS1 and Sega Saturn, which were the most popular platforms at the time. We started development on PCs using the newly released Direct X libraries - Windows 95 had just come out - but it became clear very quickly that the average PC was not capable of handling the complex 3D animal models and environments that we wanted to use. This was before the age of PC graphics cards, and the very early cards that were released just did texture rendering with the geometry calculations still left to the CPU."

 As Salvona adds, however, there was a fourth game at DMA, which would play the most influential role in SSSV's creation: "DMA were also developing Body Harvest as a launch title for the still to be released N64. The N64 dev kits were by far the most powerful hardware available in the building, so we were keen to get our hands on those and really push the limit of the hardware as much as we could."

This level is called Walrus Race 64 – no points for guessing which N64 game it was inspired by.

Power of the Ultra 64

With the team having access to the considerably more powerful dev kits, they were free to experiment with some unique ideas, including a new system for rendering the many animals which would eventually populate SSSV. "The first PC demo had a simple tiled background and a flat shaded horse animal," explains Salvona. "This test showed the potential of the Inverse Kinematics system developed by Obbe, where each animal was built with a spine and associated limbs, and deformed as it moved over the landscape or up and down slopes. To my knowledge this had never been done before on a console game in 3D. Unfortunately I don't think the 'demo horse' made it into the main game."

While ideas continued to form and development progressed, albeit slowly, further programming staff were needed and they brought aboard Daniel Leyden: "I started at DMA nearly a year after Leslie, Obbe and Grant. I had written a couple of games on the Amiga before SSSV, but didn't really know a lot about 'software engineering'. Leslie had already written a game world system and Obbe had written the code for some of the animals, while Grant had written an object system. A good foundation was in place." One of the most interesting points Daniel reveals is how freeform the design process was, with a lot of unused material: "I never saw the original design document until after the game had finished - it was very much a 'lets try stuff as we go along' kind of thing. I remember the horse which showed the Kinematics system, it was hypnotic to watch. There were certainly a lot of ideas for levels, animals and puzzles being passed around, most of which never made it into the game."

Some of these ideas would have fundamentally changed the way SSSV functioned, perhaps even warranting their own individual games. One of them, as described by Salvona, would have put SSSV alongside a multitude of vehicle customization games: "An idea we discarded quickly was that each animal would be made up of a number of different parts and that destroying a single animal would let you swap out the weapons, shields or propulsion method with your existing body. We realized this would add far too much testing time to the project - trying every combination in each level - so that was dropped."

Although the pink hippo can do massive damage very quickly, the pink hyena can paralyze it with laughter.


Although ultimately scrapped, some work had begun on the multi-limbed system, as detailed by Vermeij: "The animation system in SSSV was rather different from what was common at the time. The original design document called for animals that could change individual body parts (ie a lion could pick up and use the legs of an ostrich). This required a flexible animation system that didn't rely on [the work of an animator] but instead created animations on the fly. The code used a small number of basic walk cycles for individual legs and would use these to generate the full animation for the character. This took into account factors such as the dimensions of the animal, weight, terrain, speed, obstacles in the terrain. It was basically a flexible Inverse Kinematics system that wouldn't just modify existing animations but pretty much generate the whole animation."

The experience gained on SSSV again ties into GTA. While GTAIII went with traditional character animations created by a guy at a desk, later installments explored the methods from SSSV, as Vermeij explains: "In GTAIV the idea of code generated animation came back in a big way with the inclusion of NaturalMotion's Euphoria. This system generates animation on the fly based on the environment, physics of the character, and the 'action of the muscles' of the character. Although much more advanced than the system used in SSSV the idea is similar."

One scrapped proposal for SSSV dealt with the sizes of creatures, which would have tapped into the classic Alice in Wonderland trope, as Salvona revealed: "Another idea we briefly flirted with was that each level would have both micro and macro scales. There would be 'large' animals - mammals - but also insects on each level and you could zoom in and out as you switched between them."

Ultimately the focus was on whole body swapping between similar scaled creatures, from mice up to elephants, with players able to attack other animals until they collapsed, and then switch between bodies when close enough. At the time there had been a few, but not many body swapping games to draw inspiration from. Obvious candidates would be Taito's Avenging Spirit, in arcades and later ported to Game Boy, and the classic Paradroid on the C64. Given its popularity in Britain many of those at DMA knew of or had played Paradroid, but as Leyden explains: "Ah Paradroid, one of the best games ever! Though to be honest I don't believe it was much of an influence. SSSV was more centered around the animals differing abilities and figuring out how to use them."

Enemy bears are powerful and never retreat. It's a good idea to become a bear as soon as possible.

According to Salvona, DMA's other internal projects were far more significant, as was a sense of humor: "I had certainly played Paradroid before starting at DMA, but the biggest influences were [the original] GTA and Body Harvest, both being developed at the same time. In those games your main character jumped between a fleet of vehicles. We initially put Evo's silicon chip in as a laugh, to walk between the dead animals scattered across our test levels, but we got to like it. The notion of the game grew from there."

In a way, this positions both Body Harvest and SSSV as "missing links" in DMA's crime series, acting as a means to experiment with similar concepts to the first GTA, albeit in a fully 3D environment. Almost prototype GTAIII titles, if you will.

Play jazz with it

The slow gestation of SSSV, with time for ideas to develop and be discarded, isn't something which a lot of developers have the luxury of. DMA basically handed the keys for as-yet-unreleased hardware to a set of fresh-faced youngsters and let them play jazz with it. "The team hierarchy was very flat," agrees Vermeij. "Anyone on the team could come up with ideas and would be listened to. People, myself included, would come in over the weekend to try out new ideas. There was a real passion about the game. There wasn't much in terms of documentation so we were able to try out new ideas quickly and ditch bad ones without much overhead. There was also very little meddling from upper management to stifle creativity." What's significant is that, according to Vermeij, this spirit stuck with the team beyond SSSV. "That camaraderie, passion and the sense that anything goes was something we took with us to GTAIII and that was definitely still present when I left after GTAIV. "

According to Salvona, the company's laissez-faire attitude was due to past successes: "Most of us were new to DMA and full of blind optimism. DMA had already been very successful with Lemmings and that gave us time to explore pretty much every idea that came to mind. We had very little outside interference from DMA management. Dave Jones would appear every few weeks and note that we had new animals and levels on screen. Only after 18 months was there a push to get everything together into a finished game that could be released."

The impression given by those interviewed is that everyone helped where they could, and everyone was free to be as creative as possible. Vermeij notes how they all had a hand in everything: "It was fun to work in those pioneering days. The cool thing was that games were much simpler so you could do different things. The coders and artists would make a few levels in their spare time or record some voices for the animals. From GTAIII onwards everybody had to specialize so much that this wasn't possible anymore."

For a game on a Nintendo console, Silicon Valley can be quite morbid. Dotted around the station are dead scientists or, in some cases, just their decapitated heads!

Artist Robert Jeffrey, for example, wrote the humorous backstory to SSSV: "I was an artist for the first year of development. Most of that year was on-the-job training - like most other artists on the project. I also had pretensions of being a writer, so took the non-existent story and worked it into what it is. Although the names of the characters weren't Dan Danger and Evo - they were Broccoli and Oblong. I've no idea what I was thinking, probably placeholder names. That's what I'm saying now anyway! My other contribution was designing the exterior look of the space station. That was an experiment with NURBS in Alias Power Animator, and the design was good enough to stick for the rest of the project."

The distinctly British sense of humor, plus character designs reminiscent of TV's Wallace and Gromit claymation series, were both central to SSSV. Hardware limitations played a role in shaping this humor and style, since although the N64 dev kits were the most powerful the team had access to, things didn't always turn out as expected. "Our artists were using Silicon Graphics workstations to create the character models," explains Salvona, "and on their monitors these appeared crisp and sharp. But when we rendered them on the N64 dev boards they always appeared in soft focus - slightly smudged due to the inbuilt anti-aliasing. Someone commented that they looked almost like they were made of plasticine, so we went with a Wallace and Gromit look. Certainly Wallace's crazy inventions provided some inspiration for our level design." As for the Dan Danger character looking rather like Wallace, that's just a coincidence says Salvona: "The character of Dan Danger bears a very close resemblance to Jamie Bryan, our team lead."

Leyden described the process of transferring the work to the N64: "We had an amazing level editor written by Barnaby Dellar, which ran on the Silicon Graphics computers and allowed the level designers to do pretty much everything for a level in that one program. The levels were transferred directly to the N64 development boards which were built into the Silicon Graphics computers, and the video was output on crappy little TVs."

As supported by Leyden, the team also looked at N64 games by rival developers, since although development started in September 1995, the game wouldn't see release until October 1998: "The lead artist Jamie did a great job on the art and it had a very distinctive 'British' humor to it, which set it apart from a lot of other games. I always remember [lead programmer] Leslie and [artist] Aaron Garbut looking at every game they could get on the N64 to see what others were doing, how it worked and so on. When Mario 64 came out, that certainly made us all sit up and take notice."

Note the red blocks in the distance – as the sheep gets closer the polygon models will change to become fully detailed red foxes.

"Mario 64 was a good example of free form polygons used to great effect," adds Vermeij, when discussing the game's engine and some of the technical difficulties the team faced. "Free form polygons allow the level designer to model anything they want and ultimately the designer would use fewer polygons to get what he wants. When we started on SSSV most 3D games used a checkerboard way of defining the collision map (a heightmap where each point had a height). We did the same thing. This was a mistake, since it [had become] possible to use free-form polygons that can be any size or orientation." The team came up with a solution, though in hindsight Vermeij feels it wasn't optimal. "We worked around the issue by introducing a second layer to the map. This was clunky and awkward, and made it hard for the level designers to model the map the way they wanted." 

Although released later than intended, which had a negative effect on its marketing, SSSV benefited from being extremely polished when finally released, especially compared to some other N64 games. Along with the anti-aliasing, N64 games were renowned for their fogging – whereas SSSV had an extremely impressive draw distance. In most cases players were afforded an unhindered view overlooking the entire stage. This was a result of careful optimization of what was shown on screen. "From what I can recall," says Salvona, "the main bottleneck of the N64 was the fill rate. So although it had a Z-buffer in RAM, if you were redrawing over the same pixels again and again things would slow down. Leslie and Obbe did a lot of work to makes sure everything on screen was drawn only once. We also had multiple resolution models for each item in the game, so things in the distance used a lot less polygons than stuff up close." What makes this even more impressive is that the team had to create an entirely bespoke engine, according to Salvona: "You have to remember that in 1995 there were no off-the-shelf 3D engines. So we had to develop everything from scratch."

Leyden worked alongside Benzies on the engine and, despite his lack of experience, was determined to make it succeed: "Leslie was mostly responsible for our game engine. I remember him doing several versions of the landscape rendering over the course of the project, improving and optimizing it with each iteration. I wrote the 'object drawing' part of the game and remember being terrified that it was slow, and trying to optimize it as best I could. This was my first 3D project and I was never happy to see lots of multiplies and divides in the code! We sorted the objects by texture to optimize the drawing and also culled the objects based on what could be seen."

The draw distance on the N64 version is always impressive – it also allows players to plan ahead when solving puzzles!

Tying in with the impressive draw distance for stages is the impeccable level design and wildly diverse mission objectives, taking in a European, Ice, Jungle and Desert Zone. You can go anywhere within the cuboid realms, kill and posses any animal inside, and interact with a wide variety of objects - or simply watch them interact with each other. The team made every stage in each area feel distinct, while still maintaining a coherent style. An early level sees you using a dog to herd sheep into a pen, while later that same dog must kill rocket foxes while the sheep grows carrots in a field to be used as platforms. In one level a penguin plays music on an enormous piano, while in another a pink hyena enjoys a game of Frogger on some jungle logs. Each stage also has a hidden objective which provides a souvenir, including one where a kangaroo must punch teeth from a fossilized skeleton. Also, there are no traditional boss levels. Instead, the final level in each climate zone features something unique, including an aerial dogfight, walrus race, on-rails shooter and boxing match.

As the team explains, this diversity was due to having versatile bespoke tools, which gave everyone a free hand to experiment. According to Leyden: "Those end levels came about late in development. I believe designer Craig Filshie created the dogfight level using nothing but the scripting system. It was very impressive, with cut-scenes and enemies spawning, and really showed what the designers could achieve without having to get extra code put in by a programmer. There was a hidden game of Asteroids when you completed the game, which was written entirely using the script system, and I programmed the intro sequence using it."

According to Salvona the importance placed on good tools was part of the corporate culture. This also allowed for rapid turnaround: "There was a great tools department at DMA - still headed by Adam Fowler at Rockstar North. They built a level editor which made it very easy to create a landscape and then position items within that space. We also had Dan Leyden, who developed a smart scripting language which allowed us to define the animal behavior within certain parameters that could be easily tweaked. So an animal would run away from you if attacked but not jump off a cliff to its death, and so on. This quick feedback loop was vitally important in testing out new level ideas and seeing what worked and what didn't. Every level was tweaked, tweaked and tweaked some more before release."

The dynamic enemy AI means each animal reacts differently to those that come close to it. In this case: elephants don’t like tortoises.

Leyden added to this: "The AI was tightly bound to the scripting system, so designers could make animals do certain actions, and be guaranteed they would execute properly. You didn't want to script a bear to stand on a switch and then have the bear never quite make it! I originally started the scripting language to animate our game objects, but it soon ballooned to control pretty much everything. Barnaby created a super slick visual editor for the system. The level designers could create scripts for any object by clicking on icons and modifying parameters in popup menus. I wrote a user manual for the scripting commands - it was really easy to use once you knew what the commands did."

As a bespoke system it was flexible, though not without its issues as Vermeij recalls: "For the scripting we settled upon a system where the level designer had to pick a command from a list and then fill in each of the fields using pull down menus. This meant the game and the editor always had to be in synch - which was a pain. For GTAIII we used standard ASCII editors instead. This seems obvious now but in those days it wasn't."


 

This dogfight level, which functions very differently to the rest of the game, was specially created using the versatile scripting system. 

There's also some clever AI, with a good approximation of an eco-system. If you're playing as a small mouse but need to kill a dog or a ram, you can lure one to the other and watch as they kill each other. "For each zone," Leyden explains, "we had a grid which described how the animals responded, and the player in 'chip form' was treated just like another animal. So each animal was flagged to ignore, follow (flocking), attack or run away from each other animal. There were a couple of special cases, one I remember specifically is the king rat, which ordered his following rats to attack in force. That was a bit tricky to get working properly and not look crap." 

The flocking routines were also used on penguins, where you'll notice other penguins following you (some even declaring their love via little text boxes), while if you're playing as the king penguin, you can order your underlings to fight on your behalf. Given the impressive range of behaviours the AI allowed, Leyden was pressed on what kind of AI architecture he used, based on those described in the August issue of Game Developer Magazine: "With hindsight I would say the AI programming was state machine based, although at the time I had no idea what a state machine was. This was the first time I had written anything quite so complex; we were all pretty new and inexperienced at the time."


 

Rats will follow the king rat and attack on command. This particular level was axed from the PlayStation port.

"The animals had several high level states," continues Leyden. "Following, roaming an area, fleeing, attacking. Actual movement of the animals was controlled by another level, which took path-finding and flocking (sheep, rats, and so on) into account. The animals were all controlled using the same base functions, with a specific state machine for their individual attack moves. Attacking was broken down into advancing, performing attack, and retreating. Some animals never retreated though, like the mighty bear. When attacking, each animal would try to get to its ideal distance before using its attack move - the tortoise tank would get a good distance away, the mouse would run straight towards you."

What's astonishing about all of this is the simplicity behind it, as Leyden describes: "The entire AI only ever calculated three values for each animal: speed, direction and whether or not to perform the attack/special action. It was a huge amount of code, for three simple values!"

Aftermath and Legacy

When the game finally launched in late 1998, magazines reacted positively. EGM gave it the  Silver Editor's Choice Award, while Next Generation magazine gave it a full five stars. "The game received great reviews from critics and consumers upon release," recalls Salvona, "but as far as I'm aware failed to match the expected sales."

Part of this failure was due to an intended marketing push by Nintendo which fell through. As Salvona explains: "Nintendo worked on a scheme of dividing their marketing budget into quarters. They would have a big push for four games a year. SSSV was initially penciled in for a Q1 1998 release when it would have received the promotion it deserved. Unfortunately it slipped to a Q3 release when Nintendo put all their advertising behind their own Ocarina of Time release. Sales definitely suffered due to it being released at the same time as Zelda. Had they made their original deadline, the game would be recognized beyond just hardcore Nintendo fans, and there may have been proper in-house conversions to other platforms."

The Asteroids minigame, which eventually metamorphosed into a Defender-style minigame, created using the scripting system.


Even though it never had the impact the team hoped for, they all soldiered on, as Leyden notes: "When SSSV came out it was great to see the game in the magazines and such, however we had all been working on it for so long and other projects were on the horizon that everything moved on pretty quickly." He also emphasized the importance of a good team: "It's not where you work or what you work on, it's who you work with that's important. You could be working on a triple A title with millions of pounds behind it, but if you don't have a cohesive team that work well together to create a game they all believe in, then you will probably fail. Leslie, Obbe and others went on to found Rockstar North and create GTAIII, while DMA was eventually closed down. I stayed in Dundee and now work as technical manager at www.cobramobile.com"

SSSV left its mark on the development practices of all those involved. For Salvona, the experience helped him in the years to come: "It was an exciting time, working on cutting edge hardware and 3D techniques which were new to everyone. I learned an awful lot about code optimization and working within tight hardware restrictions, which certainly came in handy during the rest of my career. Unfortunately that has been my only games project. The games industry was fun, but I found that the hours are very long and the pay is poor compared to general programming jobs."

Purple gorillas are adept at throwing large objects and swinging on vines. They're also tough enemies to take down.

As for Vermeij, who became technical director at Rockstar North and worked on every GTA game up until GTAIV and its DLC, he feels there's some of SSSV in the popular crime series: "There is a strong connection between SSSV and GTAIII. The concept of an entity using vehicles to navigate the game world is very similar." Although no longer with Rockstar North, he still develops videogames and is currently working on www.warthegame.net

Since release SSSV has seen two ports, with mixed results, but no re-release on the Virtual Console or even hint of an update. For Salvona, he doesn't think this will change: "I would love to see a SSSV update for the current or next generation of consoles. Unfortunately it seems that while the GTA series is making millions for Rockstar North, they have their hands full."

----- ----- ----- ----- -----

Peculiar Ports

Although it never received any sequels, Space Station Silicon Valley was ported to other consoles, twice. The first was for the original PlaySation (right), apparently handled by a studio called Rune Craft. As Leyden describes it: "We were told about this but had minimal input. I believe we gave them all the models, code and textures. But that was about it."

Unfortunately this must stand as one of the worst cross-system ports of the era, with sluggish controls, massive slowdown, missing sound effects, removal of vital information related to collected items, a shockingly bad draw distance, plus entire levels missing. Salvona explains: "As far as I was told by Obbe, DMA had nothing to do with the PlayStation port. He showed it to me once on a trip to Edinburgh and it looked unplayable. The frame rate was terrible and he was embarrassed by it." The other port was to the Game Boy Color, which sensibly remade the game from scratch in a 2D format. Despite a couple of missteps with the design, which allows players to become permanently stuck in a level, it's a surprising faithful "demake".

Despite the obvious loss of its Z-axis, the GBC port of Space Station Silicon Valley manages to recreate almost everything from the N64 original.