Heavy Rain: New Screenshots and development footage*

[patsu]

*nod* *nod* I think the Heavenly Sword characters are able to capture the emotion better.

The fat detective and robber in the Market scene have eyes like dead fish's. I re-watched the HS cutscenes on Youtube. I think eye movement and exaggerated expressions made a lot of difference.

The other young FBI agent's face is too stiff.

The first Heavy Rain woman seems fine.

 

[Sigfried1977]

I think RE5 is the benchmark for believable facial animation right now. There's so much subtlety, especially in Sheva's face. If the script hadn't been written by a 12 year old, the cut scenes would have been a real pleasure to watch.

 

[betan]

I think RE5 is the benchmark for believable facial animation right now. There's so much subtlety, especially in Sheva's face. If the script hadn't been written by a 12 year old, the cut scenes would have been a real pleasure to watch.

MGS4 easily suppresses RE5 in terms of subtle details in facial animation (considering they have the exact same art direction), but in terms of conveying emotions some achievements of Heavenly Sword are yet to be matched by anyone.

 

[_phil_]

MGS4 easily suppresses RE5 in terms of subtle details in facial animation (considering they have the exact same art direction)

I'm shocked by both bolded parts.They certainly don't have the same art direction.MGS4 is in the realm of cartoney/stylized .In my opinion ,(some parts of) Nariko's(HS) and Sheeva's (RE5) facial animations are still the best yet.

 

[deepbrown]

I'm shocked by both bolded parts.They certainly don't have the same art direction.MGS4 is in the realm of cartoney/stylized .In my opinion ,(some parts of) Nariko's(HS) and Sheeva's (RE5) facial animations are still the best yet.

Yep, Heavenly Sword is still the pinnacle. And then there's U2: AT.

I think MGS4 and RE5 are both quite wooden.

 

[2real4tv]

I dunno

http://www.youtube.com/watch?v=TGPQeNWHPds

 

[Laa-Yosh]

Brutal Legend has some very nice facial animation, which is very surprising to me. I'd think it's a niche title and yet they have awesome production values (and some big name rock stars and licensed songs too).

 

[Laa-Yosh]

Oh, and for Heavy Rain, well, that's the quality you get from mass produced mocap and facial rigs... It takes a lot of human judgment and artistic skill to get it right and if you are not willing or able to put that kind of effort (money) into it then the quality will inevitably suffer. Machines and automation will only get you this far...

 

[_phil_]

yeah , they sure tried to cross the uncanny valley with a backpack way heavier than themself...

 

[patsu]

Oh, and for Heavy Rain, well, that's the quality you get from mass produced mocap and facial rigs... It takes a lot of human judgment and artistic skill to get it right and if you are not willing or able to put that kind of effort (money) into it then the quality will inevitably suffer. Machines and automation will only get you this far...

How much time does it take to tweak a character until you're satisfied (usually) ?

 

[Laa-Yosh]

I don't know how much time our guys (we have four) spend on facial animation, because we have multiple projects going on and they also work on body animations, and actually have some AA game animation outsourcing stuff on their table as well.

Then again our facial anim is from the ground up with no mocap. We don't yet have to produce such massive amounts of animation, actually our current project is the most demanding with about a whole minute of dialogue. We'll see how it turns out... but the task is definitely not comparable to what Heavy Rain requires.

However, I usually spend about 2 weeks on our most complex face rigs, the largest part of which is creating the blendshapes. This is all on top of the sculpting, modeling, and texture painting which is divided up between 2-3 guys usually.

 

[patsu]

Yeah but asked another way, how much time would you need to dress up their current characters (one of them) until you're satisfied ? Do they still have time to improve noticeably before launch ?

I can see how the eyeball animation may be a foregone conclusion, but what about the rest ?

 

[Laa-Yosh]

Ouch, that's both a very complex and a very impolite quiestion

With this amount of animation and voice work, I'm sure that the lipsync is automatic... but I have no clue about the rest of the system. I suppose they're using bones and assign coordinates for various key shapes. Like, this is where this bone should be for an "E" phoneme, this is where it goes for "O" and so on. Then you also have to be able to blend in various shapes for smiles and such so that the character can keep a certain emotive look while talking... and so on.
So it'd first take a lot of tweaking the bone rotation and position values, and of course the weighting, to see if one can get better results... But the trouble with bones is that you only get indirect control over vertex positions and sometimes with multiple bones trying to move a vertex you just can't get the result the way you want it.
Also, the common optical marker based mocap can only record a limited amount of data, for example you may only have 8 to 12 dots around the mouth. That won't be able to help you with small wrinkles or the shape of the corners of the mouth or the actual compression or stretching of the lips. That's something you can't easily compensate for with bones only and that will easily get you those very creepy looking mouths.

So all I can say for certain is that I have no clue about the work they could or should try to do there.

By the way, we're using a completely different approach, exactly because it solves some of the problems inherent in using mocap + bones for facial animation. With blend shapes we can get 100% control over each and every vertex and can even use corrective shapes for the cases where different shapes don't mix together well (like smile and open the mouth at the same time).
But the downside of this method is that it takes a lot more time to build the rig and then to animate, and it is a lot harder to automate or use mocap. I actually will have to look into some R&D on this front in the future but I'm afraid the solutions require a lot more programming skill then what little I have...

 

[patsu]

Ouch, that's both a very complex and a very impolite quiestion

It's not meant to be taken that way. I should probably replace "you" with "someone". I was just thinking out loud. My apologies !

 

[2real4tv]

new vid...watch the kids some nudity:

http://www.zshare.net/video/660801956376ae12/

 

[patsu]

Ah finally.... in the name of science !

Now what's next ?

I sure hope they have time to fix some of the criticisms before launch.

 

[2real4tv]

new interview:

http://www.gamersyde.com/stream_heavy_rain_psn_dev_walkthrough-13359_en.html

 

[patsu]

Cool, I look forward to this game even if they can't clear the uncanny valley. Am keen to show my wife the above nude scene. If you don't see me online after the game is released, you know what has happened.

 

[Kittonwy]

It's less of a problem with the character models but a problem with the facial mocapping, which doesn't seem to work as well. The character models are pretty top-notch but the way they animate doesn't seem as natural.

 

[patsu]

Missed this last week. Extended "Blue Lagoon" gameplay footage:
http://g4tv.com/thefeed/blog/post/7...avid-Cage-Featuring-New-Gameplay-Footage.html

Yum...

Looking forward to see how Mr. Cage piece everything together. Should be a hair-pulling exercise (for him !).