Computer Graphics Volume 32, Number 4, November 1998
Gaming and Graphics

Embrace Your Limitations: Cut-Scenes in Computer Games
by Richard Rouse III

"Cut-scenes" have been featured prominently in computer games for atleast the last decade, and if one looks hard enough one will notice theirpresence in games even older than that.  A good early example is PacMan, which sported amusing little interludes between some levels, featuringthe characters from the game in humorous comic sketches.  These functionedas a reward to the player for getting through N-many levels of the game,and the promise of still-more cut-scenes later in the game provided extraincentive for addicted kids to pump still more quarters into the arcadecabinet.  In some limited way, the Pac Man cut-scenes also told astory to the player, filling him in on, say, just what Pac Man and Ms.Pac Man did when they weren't eating little white dots.

The appeal of cut-scenes to designers who wish to tell stories in theircomputer games is obvious.  Instead of working in the tricky and mostlyunexplored arena of storytelling in a completely interactive environment- that is to say telling stories during the actual gameplay - designersare able to convey whatever plot elements they feel necessary through non-interactivesegments which actually interrupt the gameplay proper.  For instance,if a designer absolutely wants the player to see nasty boss monster Gargantutrontunneling out of the ground and has a really nifty effect for the same,why risk that the player may be looking at something else when Gargantutronshows up on the level?  Instead, briefly take away the player's controlof what he or she looks at, and force the player to watch a pre-renderedanimation of Gargantutron emerging from the bowels of the Earth. Another, perhaps more cynical explanation for the predominance of non-interactivecut-scenes in games is that the interactive entertainment industry is riddledwith people who wish they were working on movies, as anyone who has workedin gaming for any amount of time can attest.

But I'm not arguing for the elimination of non-interactive cut-scenesin interactive entertainment.  Far from it, I see them as anotheruseful tool in interactive storytelling.  What concerns me most arecut-scenes which don't graphically fit in any way with the game they supplement,movies which seem to have been filmed in an entirely different universefrom the one that the player encounters in the game itself.  Surprisinglyenough, this is the norm for our industry.

Methods that Lead to Inconsistency

The worst case scenario is when computer game cut-scenes are out-sourcedto film or CGI houses which have nothing or very little to do with thecreation of the in-game graphics, and are merely working from concept sketchesprovided to them, or, worse yet, vague text descriptions of what storylineis supposed to unfold in each given cinematic.  As a result, moreoften than not game cut-scenes look nothing like the art which appearsduring gameplay.  Instead of functioning as smooth transitions betweeninteractive segments in the game, the cut-scenes become jarring disruptionswhich break any suspension of disbelief the player might have developedwhile playing the game.

In terms of visual incongruity, the worst offender of all seems to bethe live-action video cut-scenes which were so ballyhooed in the gamingindustry five years ago and which now everyone seems to be moving awayfrom.  Aside from the fact that these live action segments were oftenbadly filmed, acted, and scripted, the visual dissimilarity between thegameplay graphics (be they either sprite-based or real-time 3D) and thedigitized actors appearing in the cut-scenes should have set off warningbells in designers' or producers' heads.  It seems almost inherentlytrue that filmed actors are going to stick out like so many sore body partsfrom the in-game graphics, and achieving any sort of visual continuitybetween them the and gameplay visuals is all but impossible.

One would think that pre-rendered CGI cut-scenes would have been moresuited to providing continuity with gameplay graphics, but more often thannot this simply isn't the case.  Though I'll be the first to admitthat CGI scenes have come closer than live action video cut-scenes, theyoften still look they are taking place in a realm altogether differentfrom the one in which the gameplay takes place.  This is mainly becauseartists and animators are relatively free to use whatever quantity of polygonsthey desire to create a pre-rendered scene, whereas polygon counts usedduring gameplay segments often need to be strictly limited.  And artists have natural a desire to make whatever bit of art they're currentlyworking on look as good as possible, and if they're able to use a millionpolygons in the pre-rendered cut-scenes they're certainly going to usethem, even if they can't in the gameplay artwork.

For instance, consider a game which uses a real-time 3D engine, suchas Quake.  In such a gameplay environment, artists and animators arestrictly limited in the number of polygons they can use, since the enginecan only handle N-many polygons on the screen at once.  So while ananimator might want to use at least several thousand polygons for a vaguelyrealistic humanoid figure, the game engine limits them to a couple of hundred. From my experience, few things frustrate animators who are accustomed tousing whatever polygons are needed for a piece than suddenly having touse a very small number.  But, forced (often at gun-point) by theproducer to limit themselves to only a few hundred polygons, the artistsmake due, swearing that someday they'll be able to make really swell lookingmodels once again.  And then, when it comes time to do the cut-scenes,praise the Gods!  The animators are now free to use however many polygonsthey want, since these scenes are pre-rendered on Silicon Graphics machinesand then just played back in the game as Smacker or Quicktime movies. And so the artists goes wild, using as many polygons as they want, takingtheir in-game model of two hundred polygons and increasing its count tenor even a hundred fold.  What the heck, they may throw out the modeland make an entirely different one just for use in the cinematic.  This results in high-poly cut-scene renderings which - though beautiful- barely resemble the graphics displayed during the actual gameplay. And when the player gets to one of these cut-scenes she can't help butthink (unless she's a particularly computer-graphics savvy lady), "Man,why can't the graphics in the game look this good!"

Some Games that Get It Right

But not all games are guilty of making their cut-scenes look exceedinglydifferent from their in-game graphics.  A good example of a game thatgets its cut-scenes right is Interstate '76.  Probably best describedas the Car Wars role playing game done in an arcade game style infectedwith a 1970's America sensibility,  the game includes many well-donecut-scenes which add immensely to the gaming experience.  What's especiallybeautiful about the game's non-interactive interludes is that the cut-scenesmatch visually with the gameplay graphics.  In terms of the colorpalette used, the low-polygon look all the characters have, and the stylizationof the characters the gameplay and the cut-scenes form into one cohesivestorytelling whole.  Everything looks like it takes place in the sameuniverse.

Some animators have been quick to point out to me that the cut-scenesfor Interstate '76 are not actually that low-polygon, and indeed if oneexamines the scenes closely one will count many more polygons on the screenthan could actually be rendered in real-time using the game's engine. However, to the layman who's not so savvy to graphics techniques, the sceneslook similar, even if they're technically not, and "fooling" the playeris, after all, what we're ultimately concerned with.  The game alsobenefits by using identical voice-acting during both the cut-scenes andthe gameplay, as well as having the cut-scenes lead-up directly into thegaming action, but primarily it's the consistent visual look which makesthe game a smooth experience for the player.

However, though Interstate '76's designers managed to create visualcohesion between pre-rendered cut-scenes and real-time rendered gameplay,an even better method sure to yield consistent results is the use of thegame's primary graphics engine to handle the cut-scenes.  Those cut-scenesused in Pac Man I mentioned previously are a good, though simplistic, exampleof this.  There was no technology for pre-rendered movie play-backin the early 1980's when Pac Man was released, and I'll bet that the cut-sceneswere hard-coded manipulations of Pac Man's graphics engine.  In anyevent, their graphics match exactly with those found in the game, and visualcontinuity is maintained throughout.

Some more modern examples come to us in a number of games in developmentwhich have licensed the Quake engine.  Two titles in particular springto mind: Sin and Half-Life.   These games, like Pac Man, usethe game's regular drawing capabilities to generate, in real-time, thecut-scenes the player sees.  This of course means that the non-interactiveinterludes are subject to the same polygon count restrictions prevalentduring the rest of the game, and the complexity of the scenes the designersare able to show is, as a result, quite constrained.  It's my guessthat both games, instead of hard-coding their cut-scenes as Pac Man nodoubt did, use a special, complex scripting language to govern the placementand movement of characters on the screen, as well as the movement of thecamera.  Despite polygon limitations, the cut-scenes I have seen forSin are quite well done, and prevent a wonderfully seamless continuitybetween the interactive and non-interactive segments of the game. Though I have yet to see Half-Life in action, the screen-shots I've seenfrom both interactive and non-interactive segments of the game seem tomatch perfectly, and with that game's heavy emphasis on storyline, it'sgood to know that game will present its story in a consistent visual style,allowing for the maximum amount of immersion for the player.

A pleasant side effect of using the game's engine to handle cut-scenesis that the resolution, screen-size, frame-rate, and overall quality ofthe cut-scenes on the player's screen are all identical to what the playerwill see during gameplay.  Though pre-rendered movie playback technologyhas vastly improved in recent years, and said playback gets better andbetter as the megahertz speed of the target platform increases, it doesn'ttake an expert to spot the pixelation that occurs when a movie is playingfull-screen versus the usually much sharper (if lower poly) graphics onewill see during gameplay.  Using the in-game engine, these graphicalinconsistencies go away, providing the smoothest possible experience forthe player.

An Old Pro

A designer who has been putting cut-scenes in his games as long as anyoneis Jordan Mechner, creator of Karateka, Prince of Persia, Prince of Persia2, and most recently The Last Express.  Though the first three ofthese games are arcade adventures and the last is a more "pure" adventure,all masterfully use cut-scenes to communicate their story, and all butPrince of Persia 2 use the gameplay graphics engine to render these interludes. The result in all the games is a very cinematic feel, with complete graphicalcontinuity between the gameplay and non-gameplay sections.  Not toolong ago I had the pleasure of interviewing Mr. Mechner for Inside MacGames magazine.  One of the questions I asked him was if his use ofthe game's gameplay graphics engine in the storytelling interludes wasan effort to make those cut-scenes visually indistinguishable from thegameplay.  He answered:

"Absolutely.  I think part of the aesthetic of all threeof those games is that if you sit back and watch it, you should have asmooth visual experience as if you were watching a film.  Whereasif you're playing it, you should have a smooth experience controlling it.It should work both for the player and for someone who's standing overthe player's shoulder watching.  Cut-scenes and the gameplay shouldlook as much as possible as if they belong to the same world...  [Thisis the] basic principle you have in Last Express: say you're in point-of-view,you see August Schmidt walking to you down the corridor, then you cut toa reaction shot of Cath, the player's character, seeing him coming. Then you hear August's voice, and you cut back to August, and without realizingit you've shifted in to a third-person type of scene.  Then as soonas it's over, August walks away, cut to Cath looking at August, and whenyou cut back you're back in point-of-view and now you're controlling itagain."
Mechner makes an interesting point that the player, who interacts withthe game directly, and the over-the-shoulder viewer, who watches it ashe would a movie, should both have a smooth graphical experience whileviewing the game.  Whereas one might be able to argue that a stylisticbreak in the graphics between the interactive and non-interactive sectionsmight make some sense to player who has, simultaneously with the visualswitch, lost control of their game, it makes no sense at all to someonewho's just watching the game and not playing it.

The only Mechner game that didn't use the game's engine to render cut-sceneson the fly was Prince of Persia 2, which used still-frames that appearhand-painted for its interludes.  These create a sharp break in thecontinuity of the game, emphasizing to the player a loss of control wheneverthey come up.  I mentioned this to Mechner, and he replied: "I agree with you about that.  Thereās a distancing effect to thosecut-scenes, they make you feel like you're watching a story-book. But it was the effect we were going for at the time."

Embrace Your Limitations

Recently some coworkers and I were discussing the problem of gettingour game - the forthcoming Centipede 3D - to run faster by cutting downon the polygon counts of various objects in the game world. In particular,we talked about how we could make a decent looking mushroom in less than24 polygons, since mushrooms are the most commonly found object in theworld of Centipede, sometimes with 70 or so appearing on the screen atonce.  One of them pointed out that the best way would be to havetwo pyramids - a larger one on top of a smaller one - and in such a simpleconstruction we'd have a model that, in a minimalist or perhaps even cubistway, could represent a mushroom.  I was suddenly struck by the ideathat if, from the project's inception, we had striven for a more minimalistlook, both our frame-rate problems as well as our artistic inconsistencieswould have been resolved.  Instead of having insects that tried tolook real but failed and instead looked like they were made out of (atmost) 90 polygons, we could have had insects that looked like cubist representationsof insects and which gamers would recognize as deliberately in that minimaliststyle.

At this point in the discussion I blurted out the half-joking epithet"Embrace your limitations!" which got a big round of guffaws from all present. But thinking about it later I came to see the statement as less humorousand more generally true to what we should be attempting.  As gamecreators, we need to recognize early in the development cycle what ourlimitations are, and figure out how we can make the best game while workingaround those limitations.  And if the in-game graphics are only goingto be able to use N-many polygons, and we all agree that visual consistencydemands that we make the game's cut-scenes be of a matching style to thegameplay art, we need to make the cut-scenes have only N-many polygon aswell, or at least appear to consist of as few polygons, even if we usea few more to "round off the edges."

Of course all is not so easy for the game designer who strives for visualconsistency.  What of the marketing people who, if there are fourscreen-shots on the back of the game's box, like to take three of themfrom the beautiful, movie-quality cut-scenes and only one from the in-gameart?  They would surely cry bloody murder if now all of the screen-shotslooked as "bad" as the gameplay graphics.  How would they pull thewool over the eyes of the gaming public if the game had a consistent visualstyle?  But in a perfect world, where the marketing people don't takea look at the game until it's done, I hope that we as designers and artistssee the importance of trying to maintain a visual smoothness throughoutour games, an effect which leads to the player perceiving the game as beinga more professional product.  Until the day comes when there are nonon-interactive cut-scenes in interactive entertainment, we need to makeour games look as similar as possible both when the player is interactingwith them and when she's not, whether she's playing the game herself orwatching it over someone else's shoulder.