Blooper reel from LA Noire reveals its excellent motion capture technology

(Audio NSFW)

Matt Richardson says: "These are actual outtakes from the voice actors of LA Noire, rendered out.

These hilarious screw-ups were generated as a natural side effect of the MotionScan animation process used to create the game.

Unlike other motion-capture systems that use shiny, reflective dots to track key points as an actor moves, MotionScan generates 3D models, texture maps, and meshes directly from a stereo camera recording of an unadorned actor's performance. The result is an instant virtual copy of that performance that captures mis-timed laughs, a tongue sticking out, and errant sneezes, with only minimal touch-up needed from an animator.

Ars Technica: L.A. Noire blooper reel chips away at the uncanny valley (Via Boing Boing G+ Community)


    1. The mo-cap for the screw-ups did seem a lot more real.  I’m thinking that the nature of the dialog (which seemed pretty bad) and direction probably had a lot to do with it.  When the characters are impassive cops interviewing suspects trying not to let anything slip, you’ll get, at best, subtle expressions from good actors that mo-cap of this sort is really bad at capturing.

      1. Good point actually – they’re also portraying a style of character that is kind of wooden and theatrical – which now that I think about it doesn’t lend well to a game where one of the main goals is to gauge emotion.

  1. Hilarious might be a bit…generous, here.  This isn’t exactly Burt Reynolds and Dom Deluise screw-ups or Jackie Chan out-takes.  

    The facial animation captures, though, are pretty impressive.

  2. Good start to climbing out of the uncanny valley.  The off-script discussions after the line flubs are an order of magnitude ahead of the scripted dialog on that climb.

  3. I don’t think it’s that the actors aren’t very good. The mo-cap and rendering are good enough that I could recognise Ken Cosgrove immediately – I’m so confident about that, I’m not even going to stop and look it up! The body language and facial expressions are reproduced very well, and clearly the same actor is excellent in Mad Men. So if it looks wooden, it must be part uncanny valley, and part just plain corny dialogue, that makes you perceive his performance as bad. (I’m going to look like an idiot if it’s not him.)

    1. Seems to me that it’s let down particularly by the arms.  The clothing looks like it’s rigidly attached to the skeleton, and the fingers look all weird and clumsy.

    2. Part of the problem is that the body and face have to be captured separately, so the performance is always going to be a bit “off”.

      The other problem is plain art direction: they’re effectively compositing realistic-looking video (with topology info) for the face onto an overly simple body. It looks like a man wearing a slightly oversized and stiff man suit. Better deformation and some kind of cloth simulation might have helped a bit, but ultimately the faces look like they belong in a different universe.

      Bear in mind this technology was well behind the state-of-the-art even when the game came out. There are games in production right now using capture systems that can capture facial animation and full body performance at the same time. There are also smarter “re-targeting” systems where they can map the performance from one face onto another–or, more usefully, apply whatever kind of art direction they want to the CG faces to make them fit more harmoniously with the body. I’ve seen the results and they can look amazing.

      But personally I think the performances in LA Noire will still hold up for quite a while. Just the fact they’re playing regular old people down here in the real world, not space marines or wizards, lends an air of believability.

      1. The separate capture certainly explains why the neck and shoulders on these things always appear to be trapped in a brace of some sort. My friends and I used to do impressions of mo-cap performances to make each other laugh.

  4. I think I’m going to write a modern day Philip K Dick novel called Bloopers in the Uncanny Valley:
    “Soon I realized that there were others like me…….others who had noticed the bloopers, the fluffs, the laughter…….then I realized that my life came in short, plot-heavy chunks…..”

  5. the expressions were pretty impressive.

    i wish the maker of this video hadn’t gotten so cute with the long old-school film style lead-in and the clip with the beeps between each of them.

  6. Wow, the way they shot that real video and then converted it into animation almost looks like real video!  This is the best thing since they learned how to make real chicken into chicken byproducts that taste almost like chicken.

  7. I guess they wouldn’t bother with bloopers, but for the game can they process the voices in the exteriors so that the people sound more like they are really outside?

  8. I’ve only recently started playing LA Noire and am really impressed by the level of expression that is captured. I think that these cgi bloopers have a unique authenticity to them because the technique they use allows the genuine smiles/sneezes/tongue-bites to be transplanted into the game footage without having that end-of-toy-story-2-scripted-blooper-reel feel (which I’m not knocking, but real bloopers are better). I’m loving spotting actors I recognise, like “Dead Meat” Thompson from Hot Shots.

    This is not that bit though >>>

Comments are closed.