by Michael S. Kaplan, published on 2007/01/07 18:59 -05:00, original URI: http://blogs.msdn.com/b/michkap/archive/2007/01/07/1431731.aspx
Localization can be a very diffcult task.
Now add to that the burden of trying to provide a translated version of a movie via voice dubbing and it only gets harder....
Let's look for a moment at Frankthesaiyan's efforts provide a Japanese version of one of the coolest parts of South Park: Bigger, Longer, and Uncut (the song What would Brian Boitano do?).
Here is the first version he posted, last June:
And then here is the version he posted a few months later:
Now I am not going to talk about the translation quality (I only have a little bit of Japanese under my belt, not enough to follow along with something like this!), or about the voices themselves and how close they are to Cartman, Stan, and Kyle. But whether you speak Japanese or not, you can see the biggest difference between the two videos: how well the sync-up is between lip movement of the cartoon characters and the words. Would this be a factor to consider in localization?
Looking at another Japanese version, like this one without good sync but with Japanese subtitles:
Now the queston to ponder: this can't have too much to do with suspension of disbelief since obviously these are cartoons and the voices are faked anyway. So what exactly does having the lip sync seem more real give us? And personally I find the subtitles distracting even though I have an easier time following along with my meager Japanese skills!
Of course I have trouble judging the quality of the Matt and Trey-provided translations in this other song from the movie, though I suspect that they are not accurate (those guys are mainly going for the sound, and it would be more in character for Cartman if they were incorrect!).
Thinking back to all of the foreign films I have seen and all of the dubbed TV outside of the US, there is not a whole lot of time spent trying to get the lip syncing to be perfect. Yet I don't find that to be as distracting. Am I really asking for more reality from cartoon characters than I am of actual people?
And then to bring it back to software, assuming that someone using a localized product did not need to read a bit of text that was clipped (meaning that there is no actual technical loss for them), does such clipping translate into the same kind of perception of quality issue? And would it be as annoying as the cartoon case, or as ignorable as the non-cartoon fiction case?
Random thoughts on sunday afternoon....
This post brought to you by イ (U+30a4, a.k.a. KATAKANA LETTER I)
# Mihai on 8 Jan 2007 1:26 PM:
<<And personally I find the subtitles distracting>>
I think this has to do with training/being used to it. Growing up with subtitles (nothing was dubbed), I am used to hear the original soundtrack and read the translation. In fact, I find it disturbing to see a movie where I hear the Terminator speaking French, for instance :-) And with dubbing quality is lost (a big chunk of the actor's performance is the voice and how the lines are delivered)
<<And then to bring it back to software, assuming that someone using a localized product did not need to read a bit of text that was clipped>>
It depends. If the clipping is visible, then the quality feels bad, no matter what. Imagine a dialog with 5 labels and 5 input fields, the usual grid format with 2 columns. And 4 labels out of 5 have ':' at the end, one does not. It "feels" bad.
The second thing is the target audience. Without putting people into drawers, the demands of different nations are different. What is good enough for a Chinese might be unacceptable for a Japanese. Germans seem to concentrate on the technical accuracy, while French seem to concentrate on the linguistic quality ("was written in my language" perception).
# ReallyEvilCanine on 10 Jan 2007 9:17 AM:
Avoid Germany. Not only are the dubs horrible, the voices are WRONG. So very, very wrong. And since they don't have the cultural references for many shows like, for instance, the Flintstones. Since Germans never saw the Honeymooners, they have no idea who Fred and Barney are supposed to be so they slapped them into the quickest cultural reference they could think of: Laurel & Hardy. The voices! The voices! Horrid! And the Simpsons? South Park? They're truly painful to listen to.
They've more or less butchered Scrubs as well. Germans don't understand sarcasm.
# Patrick Hall on 12 Jan 2007 7:14 AM:
An interesting statistical language modelling task... well, okay, a well-nigh *impossible* statistical language modelling task: build a machine translation system which maximizes the correspondence of "visible" characteristics of as well as all the other stuff MT systems have to approximate. So Japanese "mou" would be preferred in alignment with "more" (because those two words look similar with the sound off), etc.
I think the amount of work that would have to go into such a system would be far, far beyond that of just somehow editing the original cartoon mouths so that they spoke Japanese. ☺
And this whole topic has me thinking of Ladle Rat Rotten Hut...
http://www.crockford.com/wrrrld/anguish.html#Ladle%20Rat%20Rotten%20Hut