It’ll Be Alright On The Night: LA Noire’s Human Bloopers

By Alec Meer on February 5th, 2013 at 5:30 pm.

Mad, man

Here’s something I never thought of when playing LA Noire. Whatever its other merits and failings, the stuff it does with facial animation and performance capture was amazing, and something the whole industry can benefit from presuming it’s not drowning in a thousand million unbreakable patents (which it probably is). However, all it was being used for was to, essentially, just achieve a slightly better version of something games and especially their cutscenes already did. We can find rehearsed, scripted dialogue and, to wildly varying degrees, attendant facial emotion and animation, all over the place. What we can’t find is naturalistic, unrehearsed performances – people being people, as opposed to be people being videogame characters. Take a look at this to see how big the difference can be.

This ‘blooper reel’, in which the LA Noire tech captures actors fluffing their lines, corpsing and nattering amongst themselves as eerily accurately as it did their intoning the game’s rather more po-faced and contained script, is fascinating, infectious and convincing stuff. This is people, not actors, and I don’t believe I’ve ever seen games do that before. Imagine this, in a game. Real smiles, real laughter, real embarrassment, real mouth-fart noises.

Amazing. If this tech only winds up being used for more self-regarding GTA scripts, it’ll be a techno-tragedy.

Thanks to everyone who sent this in.

, .

62 Comments »

  1. razgon says:

    thats…kinda scary to look at.

  2. Hoaxfish says:

    The rest of their body just makes me think of Octodad

    • Treymoney says:

      Ha! Yeah! Do they not motion capture below the neck or something?

      • yrrnn says:

        The faces and the bodies are captured separately. The faces were recorded at the same time as the dialogue, using a new techniqued Team Bondi developed – instead of traditional motion capture where little dots drive a digital skeleton which in turn then drives the game/movie 3D model, the face is actually scanned by an array of cameras as they talk, creating an accurate 3D model, textures and all, of the actors face, with a new model for each frame. The bodies are just normal motion capture, and would have just been a mocap actor miming along to the recorded dialogue.

        So while the faces are very realistic – they’re a bit low res, but they look essentially identical to the way the actor’s face was recorded, including their hair and makeup – there is a big disconnect between the faces and the body, because the body motion was not recorded at the same time, in most cases, probably by an entirely different actor, and traditional motion capture can tend to look a bit clunky for a number of reasons.

        Far Cry 3 did things differently and recorded both facial and body at the same time using traditional motion capture for both, similar to what Weta did for Avatar, and I think the end result looked a lot better. Considering Rockstar/Team Bondi’s approach required the actors to sit perfectly still in a chair in the middle of a spherical room full of cameras, I doubt we’ll see this kind of capture used for full bodies any time soon.

        What I find interesting is, it probably would have taken quite lot of work to actually put these out-takes into the game and clean it all up. I guess if the face is entirely canned and not bone-driven it might be easy to just slap it onto some existing body animation, but it would still be a decent amount of work just to produce a funny video.

        • kelseypaul1 says:

          Ella. if you think Daniel`s report is something… on monday I got a brand new Renault 5 from having made $9247 this last 5 weeks and in excess of $10k this past-month. without a doubt its the most comfortable work I have ever had. I started this 5 months ago and pretty much immediately began to bring home more than $76… per-hr. I use the details on this web-site, Read about it

    • jonfitt says:

      Argh! Yes, why did they stick those realistic faces on those Half-Life 1 bodies?

      LA Noire is one of those games I’ve never seen running, just heard a lot about. I didn’t know it looked like that.

    • Magnusm1 says:

      I can only assume that in this video body-animations were not yet finished.
      I’m fairly sure that the body-animation isn’t captured by the actors.

  3. Engonge says:

    This is downright terrifying.

    • Claidheamh says:

      Yeah, talk about uncanny valley.

    • The Enchanting Wizard Of Rhythm says:

      Yeah, if this is the future of games let me be the first to say euuuurgh.

    • Premium User Badge

      Hodge says:

      I dunno, it’s the scripted bits that seem more uncanny valley-ish to me. Once they screw up and start being themselves it looks much better.

      Now I want this tech to be used for some Christopher Guest type improv stuff.

      • jonfitt says:

        I think it doesn’t help that the audio has not been post processed so it sounds like it’s in a recording studio.

        • nearly says:

          my thoughts exactly. that was what did me in, I don’t know how I’d respond to the actual game.

    • Josh W says:

      I think I must have a really small uncanny valley, this was just nice to me!

  4. perfectheat says:

    Hopefully they will be using this technology in GTA V. They are going to have to record a lot more content though.

    • Stevostin says:

      Actually LA Noire is an insanely long game. There is something like 50h of dialog recorded, which translate in my recollection to pretty much the same amount of gameplay. Took me longer to complete than GTA IV. And I maybe liked it even more, which is not a small statement. It has its flaws but I found somehow living the life of that cop over years, even by episodes, was making a better use of “the city” than the usual GTA way of dealing with it. Ideally we would have some LA Noire without the flaws (ie, without mission rating, or at least not “good answers rating”) and some procedural stuff to gather some form of ressources that would lead to improving either your transportation, weapons, clothes, housing, bank account, sexual relations and a beautiful little dog.

  5. zeekthegeek says:

    Are there John Noble bloopers? Cause I’m so in for John Noble bloopers.

    I miss Fringe already :<

    • Radiant says:

      Fringe! Fringe was nuts.
      One of the few series to say “hey! guess what… that’s right you can’t” and still be watchable.

    • sinister agent says:

      Hang on, what? Was John Noble in this?

      DOES HE LOOK SAD?

      • TheMick says:

        He kept demanding wood and oil every time you questioned him.

        • sinister agent says:

          Oh. That’s no use. “John Noble looks sad” is basically a shortcut to me sitting around in floods of tears. Fringe would be nothing without him.

          That episode set in Chinatown… crappy episode (but then a lot of them are), but the scene with Walter at the phone booth is one of the most heartbreaking scenes I’ve ever seen.

  6. stryker777 says:

    Amazing technology. Nice for Rockstar to release the blooper reel. As the article says, we see real people here, not the front that the company wants to promote.

    Thanks, Rockstar.

  7. Koozer says:

    Was hoping to see some problems with mocap making their faces do hilarious/terrifying things, but this is still quite funny.

  8. AngusPrune says:

    The facial capture stuff in LA Noire is amazing. It makes me sad to remember the gameplay was essentially “Grand Theft Gumshoe.” I really want someone to have another shot at a murder mystery type game, and this time really think about the mechanics. It’s so easy to make it something other than a point and click adventure.

  9. ulix says:

    While the facial animation in LA Noire certainly was amazing, it has one major disadvantage for the kind of cinematic experience developers usually try to create with elaborate Mo-Cap setups:

    You have to capture the facial animation seperately from the body animation, and you can only capture one actor’s facial expressions at a time.

    That means that the actors have to act out their scenes at least twice.
    Once for the facial animation, while being strapped to a chair, unable to move, and surrounded by dozens of cameras in sterile and incredibly bright light.
    That can’t be good to “get into the scene”, can it?

    And then again for the body-animations, whrer they can actually “play” or “act out” their scenes like they would on a stage or in a film.

    Another “disadvantage” (if you want to call it that) of the tech is of course that your virtual characters have to look exactly like their real-world actors.

    I much prefer the approach Naughty Dog or Rockstar (this being Team Bondi) are using, where they only capture the performances of the actors once, face and body. The facial animations might then not be as detailed or realistic (and usually need a lot of retouching by animators), but I think the overall performance you get to see and hear is more convincing.

    • TechnicalBen says:

      They need a Kenect style camera (multiples) to do the recording while also doing the mocap. Record the entire actor in 3d. Probably need way more resolution and laser scanning than it’s worth though. Plus the point cloud data for the body would be prohibitive at a guess.

    • Premium User Badge

      FriendlyFire says:

      Yeah, fully agreed. While LA Noire’s face capture and animation tech is bewildering and the faces are truly lifelike, I think the overall acting and especially movement feels a lot better in Naughty Dog’s Uncharted. Uncharted is one of the few games where characters feel like people, even if the face capture isn’t as good.

      LA Noire’s faces are let down by the rather stiff, underdetailed bodies.

  10. pupsikaso says:

    I don’t know… this is just short of falling into the uncanny valley. Any more realistic than this and I would be more horrified than amazed. I’m already a bit horrified at times when watching this.

  11. Senethro says:

    Why is this thread a bunch of dudes going “eugh no” instead of “How would I use this?”

    • gritz says:

      Seriously! It’s depressing to see people dismissing such an awesome technology because of the uncanny valley.

      • terry says:

        Awesome tech it is, but also deeply flawed and labour-intensive. I’d be a lot more enthused if it was within reach of the average developer, but it ain’t.

        • gritz says:

          Labor intensive I get, since it uses actual actors instead of voice actors. But deeply flawed?

    • The Enchanting Wizard Of Rhythm says:

      ‘Hey this looks terrible, I wonder how we could use it’- No one ever.

      • spedcor666 says:

        ‘Hey this has potential, I wonder how we could use it’ – Not The Enchanting Wizard Of Rhythm.

      • sinister agent says:

        The animations are incredibly good. It’s the uncanny valley effect that is bad (although less so than many others). So, use the same technology and alter it slightly for use with less photorealistic models (think TF2 or somesuch), and you get the great animation without the creepiness.

      • gritz says:

        Except it looks awesome and only creeps out people who have uncanny valley hangups that they should work out on their own.

        • Ruffian says:

          Yeah, idk. I’ve seen a lot of animation/models that did give me those uncanny valley “heebie-jeebies”, but this one didn’t quite do it for me.
          The raw voice tracks certainly don’t help things much. I could definitely see how the roughness of it all could coalesce into some kinda creepy uncanny gumbo for some. Being someone who does some occasional home recording though, I think my brain kinda “normalizes” stuff like that fairly quickly. Really interesting to see this sort of stuff for games, regardless.

      • Slurpy says:

        ‘Hey this looks terrible, I wonder how we could use it’- Jersey Shore Producer

  12. Tei says:

    Is like everything the article say. I don’t remember seen a NPC tongue before. These things are alive!.

  13. Dowr says:

    3:58 That is bloody creepy.

  14. Mario Figueiredo says:

    Incredible and amazing. It really raises the bar for actors in games. Their contribution to a game becomes now more obvious and real, which puts a whole lot more pressure on gaming studios to hire the right people. Gamers can only benefit from this type of technology.

    Unfortunately Rockstar is’t a company I’m interested in. Their games never drew me. Their themes never interested me. As great as they games seem to be, the problem here is really I don’t connect with murder cases, street gangs, escape convicts and most definitely with anything even remotely associated with the American Old West.

    And here the fact Alec touches on patents is an eye opener to any of us even remotely interested on this technology. Because even if you like Rockstar, this sort of stuff would be incredibly useful everywhere else. Here’s to hope any patent (if existing) won’t be able to stop the development of alternate similar technologies by other studios. Or at least that a license mechanism is made possible.

  15. smoke.tetsu says:

    As far as uncanny valley stuff goes this is the least uncanny valley thing I’ve personally seen. Even the stuff in Tron Legacy with young Jeff Bridges was more uncanny valley than this in my opinion or… the CGI Orville Redenbacher or most of Robert Zemeckis’ CGI stuff. Also I quite enjoyed the game myself.

    Also, one complaint I see people make against the graphics in this are the clay look of the faces. Well, I think the problem is pretty much the low resolution of the final output.. like the texture maps, normals, etc. for the faces blurring the final output. If they up the resolution of that it may help. Although seeing it more clearly may make things worse for some people I guess.

    But yeah they basically made a fully 3D video for the faces with multiple cameras surrounding the person but then either down-sampled it heavily or captured it at a sub-HD resolution.

    • Mario Figueiredo says:

      Indeed. I understand that the Uncanny Valley is different for every person (but not quite, see below). But I’m starting to fear the term has entered into that nasty bog of overuse and relative meaning.

      I don’t think Mori meant for it to be reflection of individual perception. It’s more of an effect that is linked to cultural and social aspects that can be shared by a group of people.

      It’s hard to attribute to an Uncanny Valley that Joe is more and John is less disturbed by an image. That’s to do with individual experiences, education, traumas, etc. It’s the matter of psychologists and others. The Uncanny Valley reflects on a whole society or a large group of people. It’s easier to attribute to the Uncanny Valley that Joe and Jane look differently at an image, one being man and the other woman. Or that Joe and Musharaff feel differently about that image.

      So I think the term Uncanny Valley should be used with a whole more moderation.

      • Ruffian says:

        I get what you’re saying (i think, I mean I realize the term is supposed to refer to a certain point of almost realism that evokes disgust for a majority of people) but what term could be used to better summarize that odd feeling of disgust when something looks disturbingly human, but at the same time, not, on the individual level? I mean sure there could be plenty of other psychological issues affecting a body’s perception, but wouldn’t you be able to tell, to a degree, that it’s because it looks almost human? Or is the reaction supposed to be a completely unconscious thing?

        I’m really not being dickish, I’ve just honestly never heard of any other term linked to this specific sensation. Again let me apologize if I’m coming completely out of left field here, or understanding you wrong, I’m something of a layman in the area. I mean I’ve read the definition, ofc, just never really researched it in any real depth.

  16. Turkey says:

    Man, I can’t wait to see the realistic looks on their faces as I mow down a billion terrorists in Codblops 26.

  17. Phantoon says:

    This is amazing, and hilarious.

  18. faillord_adam says:

    Was I the only person constantly seeing ‘Steam boner’?

  19. Premium User Badge

    Ninja Dodo says:

    The technology remains impressive… but I would like to see facial animation this detailed that’s *dynamic*, not captured… because this tech is inherently unsuitable for anything non-linear. They’re only ever going to capture a finite set of takes.

    Imagine auto lip-sync and layered emotes that approached this level of detail, either built on captured elements, or hand-animated so they can be isolated and easily layered on top of each other. I think Source is still the closest thing to this, but I would like to see someone make a game like LA Noire that has more room for the player to actually experiment with conversations, rather than One Correct Answer for everything.

    I was actually disappointed that LA Noire was not more like Phoenix Wright in that regard. Even though that was still linear you could at least try different things before arriving at the solution.

    Of course, even if you solved dynamic facial expressions you’d still have the bottleneck of voice acting.

  20. Michael Fogg says:

    The hell you people are talking about, if you want to see uncanny valley check out this video
    http://www.youtube.com/watch?v=UoPplpBPQxQ
    Modern day computer charachters are not particularily creepy at all, I think some pundits are greatly exaggerating the effect.

    • Premium User Badge

      Thermal Ions says:

      Yes, also don’t get the Uncanny Valley people are going on about. The LA Noire video actually just got me thinking I should finish the game one day.

      That Daft Punk video – now that’s creepy.

      • Vurogj says:

        The Daft Punk video isn’t uncanny valley, because at no point is the robot anything other than an obvious robot. Uncanny Valley is when something looks human but isn’t, and the way the heads and bodies don’t behave properly with respect to each other in the blooper reel is UV.

        Creepy I’ll definitely agree with, but not uncanny.

        • LennyLeonardo says:

          Yeah, the thing with Uncanny Valley is that the effect is meant to be more pronounced the more realistic the representation is. Think Stepford Wives, or George Osborne.

    • Ragnar says:

      That video isn’t illustrating the Uncanny Valley – it’s simply a very creepy robot.

      Some people are getting uncanny valley vibes from LA Noire because the facial animation is very lifelike and human, but trapped within low res faces and stiff bodies that make it feel not lifelike and not human, and thus the conflict.

      However, rather than making it less realistic, you could also make it more realistic and climb out of the valley that way. Higher res images, better animation for the bodies, post-processing on the audio, and you climb out of the valley on the human side and all is well.

  21. The Random One says:

    Amazing tongue technology here.

  22. vondas says:

    Oh my God, this is hilarious.

    I really don’t know why people find it creepy though, it seems fine to me.

  23. E_FD says:

    The sad part is that this looks so much more naturalistic than anything in the released game, where people’s mannerisms involve lots of forced, artifical blinking and eye-batting to show that they’re supposed to be lying. The D-list cop show dialogue didn’t help either.

  24. Stevostin says:

    It looks a bit odd and I remember initially not being convinced at all with the tech. Thing is once you’re used to it, it’s not odd at all and it really brings a lot to the story and the gameplay. I loved reading faces and use it to make a guess.

    Also… So much more content! This game was incredibly generous. It could have been twice shorter and still seem a long game. It’s really the equivalent or one, or even two seasons of a good tv shows.

    Also story… music… atmosphere… Neat neat game. As RPS said, probably the most ambitious Rockstar Game. I think their main issue was to inform the player what was a winning and a loosing answer. Letting it all undecided and not rate it would have make it a much more enjoyable experience, I think, freeing you of stupid temptation to go over again a whole sequence just to get the “right” answer.

  25. elhreno says:

    Robophobes…

  26. Nixitur says:

    While interesting, realistic facial expressions do not make a good cutscene or a good movie, just like real speech is thoroughly disappointing and makes for horrible dialogue in basically all of media.
    The worst thing an interviewer can do to the interviewee is printing -exactly- what was said, down to every little stutter, pause and messed up sentence. Same thing with facial expressions. As amusing as the video is, those are special circumstances and having real facial expressions throughout the entire game would be rather annoying.