[What happens when developers lose control of a nearly-finished game soon before its planned release? After forming a new studio, Runic Games, many of the developers behind Mythos transitioned to Torchlight. What follows is an examination of the decisions, steps, and key design elements that have shaped the look of Torchlight -- a title produced on an aggressive schedule and modest budget that still strives to for the production values of a big retail game.]

Our Mythos

It is impossible to tell the story of Runic Games and Torchlight without covering what brought all of us here: a little network test that grew up into a game called Mythos. It began as a one-man "team" of current Runic president and lead programmer, Travis Baldree, working with the Flagship Studios team to create a small hack-and-slash game as a network test for Hellgate: London.

Accelerated progress and increased interest in the test (at the time labeled "Project Tugboat") allowed Travis to expand the team to three. Eventually three became five, and then eight (which is the point I joined.) Ultimately, we had a small, focused group of 14 at Flagship Seattle working on Mythos. The scope had changed from a simple test to plans for a free-to-play MMO with persistent areas, a cash shop for items, and a variety of features considered standards in the MMO space.

Mythos had a strange aura about it. It was a fairly straightforward ARPG, very much in the Diablo mold, with a likeable look and easy-to-pick-up gameplay. It wasn't groundbreaking, it wasn't a technological marvel, nor was it particularly shiny and beautiful -- but it absolutely was fun.

We opened it up to Alpha and Closed Beta testers and something clicked. Testers and forum users were having fun, making suggestions, and behaving like -- dare I say it -- reasonable people, with nary a flame war or troll to be seen. This atmosphere was a bit odd to us, as we knew all too well how the tone differed in the forums for the Hellgate: London team down south.

But we ran with it, and reciprocated with constant interaction -- and as thin a veil you could imagine between our testers and what we were working on, contemplating, or fixing. This aura of being in synch with our game's audience was something special; they appreciated our openness and we appreciated the input, enthusiasm, and tone in the forums. We were a few weeks from going to Public Beta and it was all going so well...

Then, in July of 2008, Flagship Studios closed and the Mythos IP was lost in the fallout.

One month later, the Seattle team, along with Max Schaefer, Erich Schaefer, and Peter Hu, formed what would become known as Runic Games. Having been so close to releasing our game only to lose it all on the home stretch, we set forth with an eye to finish what we started... by starting all over again.

The Elephant in the Room

Just as it's impossible to discuss where we are now without mention of Mythos, we're likely also going to be linked by many with the benchmark for ARPGs: the Diablo series. Comparisons to the upcoming Diablo III are inevitable -- so let's get this out of the way.

Our founders, along with Runic's composer Matt Uelman and QA Lead Ian Welke, were major contributers to Diablo and Diablo II. Their expertise and enthusiasm for this type of game has been absolutely invaluable in creating Torchlight. It's one heck of a foundation to build a studio upon.

Any fan of ARPGs is aware that Diablo III was announced and is in production. All of us here are fans of the genre, and are very excited to have a new Diablo game to play. We've also already heard comments that Torchlight is trying to "take on" D3 as direct competition -- or even that Torchlight is a "cartoony version of D3".

Both comments are actually quite incorrect. If anything, we've actively proceeded knowing full well that D3 is on the horizon and that it behooves us to be on a different train track when it comes roaring by. Certainly the ARPG, hack-and-slash gameplay full of loot drops is absolutely a common ground, but we're producing a single player game as a launchpad to a free-to-play, hack-and-slash MMO.

Gameplay similarities aside, the art controversy surrounding the announcement of D3 was somewhat fascinating to me and, in a small way, contributed towards Torchlight's look. My reaction to the first screenshots was that Blizzard had created a beautiful, living painting. To me, it was absolutely gorgeous, and a style that I would have loved to have pursued.

But knowing the direction Blizzard went with the game helped inform our decision to go in a different direction. We didn't want the inevitable comparisons, so in a way... it helped us define which artistic choices were now off limits.

Proper Preproduction

As we began the project, Runic Games had a great team, an idea of the type of game we knew and wanted to make (again), and initial funding to allow us to create a prototype and pitch our ideas to publishers.

What we hadn't completely anticipated was just how much buzz Mythos had generated and the subsequent doors it opened. We quickly found ourselves in the position of having a number of options, ranging from creating a similar style game with some well-known IP, to total acquisition. During this "deal making" phase we kicked off preproduction and experimented with possible routes we could pursue.

This was about as different a beginning as we could imagine to what happened with Mythos. With Mythos, an art style had been set, the game was well underway, team members jumped in and hit the ground running, and we were very much laying down tracks ahead of us as the train was in motion.

With Torchlight, we have been able to strike a far better balance between design and production. It seems quite obvious, and even silly, to say, but preproduction is absolutely invaluable for any project.

Often, and especially with new studios or smaller teams, the push to get into production right away is a difficult force to resist. Sometimes, the deal with a publisher is too aggressive from a schedule standpoint; other times it's a budgetary concern; and sometimes you don't have the team fully assembled yet.

For us, we knew what our deadline was to solidify a deal, and we made the most of our time leading up to that point by creating a prototype, concepting the necessary elements of our game, and creating the tools that would make production as efficient and painless as possible.

With the options that were presenting themselves, we originally targeted a "very casual" look for our game. One of the original ideas was to go even further simplified and inviting than we had done with Mythos. From a practical standpoint this made a lot of sense and we put together some initial concepts for pitches.

The original and "very casual" Delvers pitch image

This look addressed some of the interests of our possible deals, but ultimately it had no champion. However, the idea of using the softer, more painterly environments began here. There was a little something rich and charming here, but the characters left us cold. I think the overwhelming feeling was this was "too kiddie" for us.

We quite literally went back to the drawing board and tried out a few more ideas. One proposal that was pitched was using a unified body mesh with interchangeable heads to create our playable races and genders. The primary benefit was to our schedule, and the simplicity of the art pipeline.

At this point, we were still exploring various game engine options and there was a push to keep things as simple as we could. The art team explored this "unibody" design to see if we could get all the character and contrast we wanted out of this type of construction.

We tried the "unibody" test with some classic fantasy races

The "unibody" test gave us something we enjoyed more so than the original pitch, but we were now faced with the question of "is this a good enough tradeoff?" What's the value in sacrificing our desires to do unique playable character meshes worth the simpler, more schedule-friendly approach?

This was debated by the team while we whipped up a prototype character model based on the unibody sketches. We liked some things about the model, but it became increasingly obvious that the desire of the art team was to go in a different direction. We began to contemplate ways we could create unique meshes for our playable characters without a pipeline that would jeopardize our timetable.

The "unibody" prototype helped to push us toward our eventual designs -- by pushing us away from this, toward something else.

The texture style for our prototype was too flat and unrefined. However, it had the basic principles we wanted: a combination of simplified shapes with inked line work, and comic book-style gradient coloring.

These ideas would survive into our further work, and evolve into something more visually appealing.

The simplified shapes and angular stylization of features at this phase were too simplified, but we felt the general idea was worth pursuing and could be improved upon.

The primary issue for all of us was that the basic proportions for the male characters didn't feel heroic, and the females didn't look feminine. The heavy stylization of the characters overshadowed this.

Once we all realized that we would need at least two body meshes (a male and female) to make this work, the jump to three meshes was not so daunting. It was about this time that we had settled upon the idea of doing a single player version of the game prior to the MMO version.

The decision to present three defined classes in the single player emerged and it dovetailed into our character design problems perfectly. The team decided to create three completely unique characters to represent our three classes, and defer a highly customizable playable character setup until the MMO.

Coming to this decision early on was vital. When we were stuck in the "unibody" mindset, morale dipped, as some felt art was being sacrificed for schedules, and that we were trying to shoehorn in a solution. Once we made the call for unique characters, the mood changed considerably and we were back on track.

Similar principles, but with a more heroic structure

The Prototype and Our Design Process

Once we were happy with how the initial character designs were progressing, we put all effort into the prototype of the game. Along the way we had decided to utilize the Ogre 3D engine and already had a bare-bones shell of a level with some art stubs to run around and smack things.

We were still heavily into the "deal-making" phase; a few options had risen to the top of our list. As a studio, we put all effort into presenting something indicative of the game we were trying to make and getting it ready for outside eyes.

The first step towards doing this was putting together our first tile set as a benchmark. To work out some of our fundamental construction problems, we went for the obvious dungeon type: the "dark, gothic, undead-filled crypt", for its ease and architectural simplicity.

As with Mythos, Torchlight was going to utilize random dungeons, but we were eager to do it in a different way. We had found the Mythos solution of smaller tile pieces with rule sets for how they connected to be too limiting. Often, dungeons felt like an extruded crossword puzzle -- and very similar no matter how random the rule set was.

With Torchlight we opted to build "super rooms", or "chunks", that had constant attach points where we could link any of the other chunks for that tile set... and within these chunks we could construct and decorate with more purpose and focus. We could lay out actual spaces with a coherent theme, rather than hope that when we randomly rolled for props, the layouts we ended up with worked out.

There was also a new sense of verticality we didn't have in Mythos. It was absolutely the correct choice and our Torchlight dungeons put our previous ones to shame.

The crypt level -- from a rough thumbnail, to a cleaned-up slice, to the final level in-game.

Runic Games has a fundamental belief that "you always have and maintain a playable build of the game, all the time" which is directly attributable to Travis Baldree's previous projects. It was with this philosophy that we took our shell of a game and alternated between playing it and iterating on the various elements.

As a group we're probably outside of the norm when it comes to design process. Our structure and work culture is designed around the idea that we're a team striving for the same goals and that every single team member should feel like they can share ideas or criticisms on any portion of the game regardless of their role. Our offices are of an open design, where everyone can see everyone else and issues that might warrant a meeting in other studios are just a quick conversation away for us.

We are often asked how we're able to do so much in such little time. Besides the fact that Travis and others on the team do the work of three or four people, our structure and methods for essentially circumventing a traditional meeting unless absolutely necessary is a big part of this.

There is also much to be said for hands-on, practical, iterative game design over trying to solve things hypothetically within a design document. We're not completely dismissive of design docs, but in documents can be held with too high a regard. Sometimes it's just a matter of implementing it, playing it, tweaking it, and playing it again until you have the results you want. Between the Diablo games, Fate, Mythos, and now Torchlight, our team has much experience with this type of game.

Concepts and rough potential variants for a creature in Torchlight.

Honing the Style

When developing the overall style of Torchlight we drew from our own experiences with Mythos while tried to address some of its failings.

For example, we incorporated a new approach to texture styles. Our team shares some design sensibilities; improving while playing to our artists' strengths, when possible, was logical.

Once we had the deal in place with our publishing partners at Perfect World, Runic Games was in a unique position of being able to use our single player game to develop IP that would build up our universe, lore, and inhabitants for the following MMO.

Establishing what people could expect visually from a Runic Games title has been a prominent factor in many design decisions. Like any studio, we want our art to stand out and be recognizable.

In approaching Torchlight, it was necessary to analyze what we had just done with Mythos and learn from it. The comments on Mythos' style were that it was bright, colorful, and inviting... but also somewhat generic and, well, "WoW-ish".

In hindsight, I would agree. However, few people saw our third zone in Mythos, which was the first zone that didn't use a single piece of the early, inherited, outsourced art... and it was this style we wanted to push even further with Torchlight.

Alchemist gear subset. It highlights our bold, "chunky" trim and design elements.

When describing the visual style we're pursuing we often talk about "chunky" details: using simplified shapes, streamlined, bold features, and giving things a sense of weight. We also talk about "clean" versus "loose", and how you can create separation between elements that are both aesthetically pleasing and also improve gameplay.

We cover a lot of the fundamentals, as any art team does: silhouettes, color palettes, and the like. But we always return to a design being "fun". If it's a static and mundane object, does it have character? If it's a weapon or piece of gear, is it something we want on our own avatars? If it's a monster, is it something that's going to be a joy to slaughter?

What are the monsters' motivations and how does this impact design and animation? Is a design memorable, and does it reinforce our goals for the game? All of these ideas contribute to the style of the game -- but there is far more.

The summonable Alchemist imp: with chunky proportions, and full of quirky personality.

With Torchlight, we're also after a tactile, visceral, and emotive style that both supports and is supported by our visuals. How things interact, the way a simple attack feels to the player, the feedback from a perfectly executed audio cue, the mood of the ambient sounds and the musical score -- these are all elements that, when done correctly, are interwoven with the visual style. Breaking this relationship down is best left for another article, but it's important to at least recognize this basic idea.

Two Texture Styles and the Importance of Contrast

One of our goals for the game was to have an incredibly low minimum spec so that we could reach a very broad audience.

One of the results of that decision was to do everything without shaders; we didn't want a small subset of our users to be seeing the game as intended and the majority of users getting something else.

When we opted to go totally fixed-function, we began exploring a more "artistic" take on texture styles, to not just differentiate our look, but also to try and keep things visually interesting in a hand-painted sort of way. We worked primarily with diffuse maps with spec and self-illumination options.

This choice also contributed to our straightforward art pipeline and allowed us to create a tremendous amount of assets and in a relatively short amount of time.

We had enjoyed the softer environmental tests we had done early on, which had a very painterly appearance, and we also thought a crisp, "comic book coloring" approach that we attempted on early character prototypes was worth pursuing further.

We liked the theory of classic animation: backgrounds which were soft and painterly, combined with moving elements with crisp outlines, a simplified structure, and clean, layered gradients for coloring.

Crisp and clean character textures versus loose, painterly, faux-watercolor environmental textures. This combination helps lift our characters and gear from the backgrounds.

I had desired to take this approach with textures for some time, but Torchlight was the first project where it felt like a good match. It was primarily a matter of hitting that look and finding a workflow that maintained consistency.

There was an adjustment period for both our internal artists and our outsourced team of artists to become comfortable in this style. But, once everyone had the style down, it proved to be a relatively fast method of texture painting -- which, admittedly, was a factor in pushing our team to attempt it.

This "burrower" highlights our texture style and shows our desire to take familiar things and present them in a different and interesting way.

The Tone of our World

The number one question we get asked regarding our art style is "why did you decide to go this route instead of something more realistic, dark, and gritty?" The answer is pretty simple but there are a few considerations.

The simple answer is that we're making a fun adventure game, not a bleak, gore-filled, depressing one. We want to present an inviting world that is fun and thrilling and create sections that are more grim and sinister in contrast to that. If the entire world is bleak and grim, it starts feeling samey and the impact lessens over time. We want to present interesting areas, creatures, characters, and storylines that support a fun adventure world first, and sprinkle in tonal changes in a more calculated manner for dramatic effect.

We want to create the single player game to set the tone and help establish the world. Torchlight is intended to be an inviting world for people to hang out in and explore. As we shift into the MMO, the tone can shift; dark elements can be introduced without suffocating players. Variety is the spice of life, as they say.

A sample of our whimsical designs. Remember, "fantasy" is the antonym of "reality".

Secondly, going realistic, dark, and gritty is fine for many projects. But because of that, there is no shortage of these types of games; the team here enjoys many of them. As a new studio with the opportunity to create our own IP, though, we wanted to stand apart and define our own look -- one that the entire team has a real, genuine interest in.

As a group, gritty realism isn't our thing. We're a whimsical, quirky, and playfully twisted bunch whose natural tendencies in art reflect that. We'll leave the gritty realism to the teams with a passion for creating that look.

The challenge for us has been finding the perfect balance that is inviting to new players, but not off-putting to fans of the ARPG genre. It's a tough line to straddle, at times, and some may argue that you cannot please both.

This may be true, but with Mythos we found that we got very close to that blend. We had hardcore players and absolute newcomers all enjoying what we were building. Our team learned from that experience by building upon our experiences, addressing our shortcomings, and adding in new ideas so that Torchlight would surpass it; we hope that it will do so in every measurable way.