Naughty Dog lead game designer Richard Lemarchand capped off the Thursday schedule with an hour-long discussion of the long development process behind the character animation in Naughty Dog's critically acclaimed action-adventure game, Uncharted: Drake's Fortune.
Throughout the talk he stressed the importance of planning, yet not so much as to limit improvisation, and warned of the danger in getting hung up on doing things the "right way" as a matter of principle, when often simpler solutions and workarounds are the wisest course of action.
Above all, Lemarchand stressed at every step of the session the importance of close collaboration throughout the development team, on all levels, across all disciplines.
Uncharted had about a three-year development cycle; a year of pre-production, followed by two years of active production. Early on they began to research all manner of pulp adventure fiction, from Tintin to Doc Savage, to seminal movies like Gunga Din and more recent stews like Raiders of the Lost Ark and The Mummy.
Beyond the hair-raising, larger-than life quality of these stories, the team wanted, wherever possible and appropriate, to capture the "certain lightness of tone" in the source material, to contrast with the current standard for Western games, which Lemarchand described as "overwrought and all a bit emo."
Cigars of the Pharaohs
From there, the team laid out a set of gameplay themes that seemed to tie in with the scenario: traversal (that is, the logistics of working one's way through an environment – jumping and climbing), exploration, gunplay, and hand-to-hand combat. As in the pulp roots, gunplay was to be largely "cover-based", tying in with the idea that the protagonist is not so much a hero as an average guy in way over his head, pulling through more out of determination and luck than any special skill or bravado.
For a while, they assumed the game would need an auto lock-on feature, to keep up the driving pace and deal with the complexities of third-person aiming. The problem was, auto-aim didn't feel very visceral. Beyond that, it was not particularly challenging or interesting, making the inclusion of gunplay at all seem a bit pointless.
Then Resident Evil 4 appeared, and everything seemed to fall together for the development team. Using various camera transitions, they tried their best to eliminate the feeling of different "modes" to shooting and exploration, though the process was somewhat harder than they expected.
Regarding character and facial animation – which the team considered one of the most important elements in expressing their desired tone – Lemarchand said they "should have got basic [systems] in earlier"; instead, they "kept trying to do complex things," accounting for every potentiality of interacting with the environment, "that held us up." "Get on," he advised. "Build the simple stuff first."
Throughout development, animators and programmers worked literally side-by-side to get the results everyone wanted. "Close collaboration is key," Lemarchand said. Different disciplines should not keep team members from interacting; in fact, every discipline needs to be informed by every other discipline, to create a coherent vision.
Sketching in Motion
After the team had a rough idea of tone and premise, they knocked together a few concept sketches, storyboards, and pre-visualization cinematics. The initial "pre-vis", which showed the protagonist dangling from ledges, leapfrogging and diving behind cover, and awkwardly shooting a pistol helped to inform all future development; most of those actions and behaviors were incorporated more or less directly into the final game.
Other, contextual animations – where the protagonist notices and reacts to things around him, rather than just robotically running through (for instance) a running animation – were improvised at the motion capture stage. Subtle facial animations were done with keyframe animation.
The original design of the protagonist was much lankier and odd-looking than the final version; in his initial design, much of the production team had trouble getting into his head, and understanding where they were going with the character. Rather than battle for the original design, he was changed to resemble (in looks and attitude) Johnny Knoxville, from Jackass. Once he was changed from a goony young man to a "dude" who radiated "cynical humor" and "coolness", everyone was on the same page.
The Next Handhold
The E3 2006 trailer, as with the pre-vis, became something of a touchstone for future development. Their goal became to make everything in that trailer possible within the actual game – and for the most part, Lemarchand says, they succeeded. After the trailer, though, they realized their tool set was a mess.
Once again they had spent far too long trying to account for everything, resulting in a buggy heap that nobody could use properly, that was frustrating the whole team, and was bringing development down to a chug. So they chucked most of the new system and went back to a more advanced variant of the Jak tool set. Again Lemarchand stressed, "Don't try to solve everything. It's nice to aim high, but choose your battles. Workarounds are often fine."
In an unusual step, the team chose to also use the voice actors for at least some of the motion capture work – including all of the cut-scenes. This resulted in a certain amount of cohesion in the performances, and a good deal of improvisation that turned out better than what had been scripted. "In fact, some of the best moments," Lemarchand said, came from this improvisation.
Cut-scenes were done in-engine, yet not real-time. "Not to cheat," Lemarchand said; rather, out of concerns for load time. It was quicker on the player's end, for the game to just load and play a movie. Cinematics were lit "not even just scene by scene, but shot-by-shot", resulting in a highly dramatic, deliberate sense of staging still rather unusual in video game lighting and camerawork.
The Power is Yours!
As with the pre-vis and trailer, the E3 2007 demo was yet another point where everything came together. After running through the reason why demos are "a double-edged sword" ("they really suck to do"/they can bring the team into focus), Lemarchand mentioned how he appreciated the "constructive negative feedback" they received. He noted that they create "very little throwaway material" for demos; everything is pretty much sliced out of the final game plan, in chunks.
Wrapping up, Lemarchand went back through the points. "Timely scoped planning is important, but don't over-plan... Getting on with making the game is the best way to make it." Then he stressed, above all, the importance of "a flat hierarchy, an open-door policy, and a meritocracy"; having a company culture based in "hard work, teamwork, and mutual respect."