Beyond: Two Souls

[patsu]

I'm still early in the game.

Based on the Asura demo I played a while ago, there are some subtle differences between the QTE concepts.

There are essentially 2 kinds:

(A) Context sensitive buttons. These appear when you're near objects that can be interacted. You can take your time to use them.

(B) QTE. These need to be completed within a short time. In B2S, a few of these events don't really have a clear on-screen cue. e.g., In unarmed combat, you move the right stick in the direction that feels natural (blocking an attack, or dodge one).

You can control the character to explore the environment.

EDIT: Argh, I can't play this game more than 30 minutes without getting dizzy. >_<

 

[RenegadeRocks]

EDIT: Argh, I can't play this game more than 30 minutes without getting dizzy. >_<

Good for you guys who can enjoy 3D without headaches ! There should be something that troubles you guys too ! I can't enjoy 3D at all, most of the time I can't see the depth (looks flat or double vision to me) at all and end up with squint eyes, or the feeling of that.

Good to know there's something that you 3D enjoyers can't enjoy ! The punishment for playing the game before I could ! Serves you well ,Burn !

Spoiler

 

[patsu]

I adjusted sensitivity down to 30. Will try 10 tomorrow.

 

[Sigfried1977]

The game supports 3d?

 

[loekf]

The game supports 3d?

No, 3D display tech is dead until the industry finds a way to do this w/o glasses, headaches and dizziness.

I also finished the game. It's around 10 hour game play, so not that long. The final levels with lots of action are truely amazing. You can't influence a lot, but it was quite a ride. It still more intense that e.g. watching a Holloywood blockbuster with lots of CGI in your living room on Blue-ray ;-)

 

[Sigfried1977]

I personally preferred the more laid back chapters. Cage's blockbuster endeavors are just as stupid and ludicrous as they are in the CoD competition he so despises. Heck, the Somalia chapter is hammier and more stupid than anything CoD has ever done.

 

[BoardBonobo]

The whole Somalia chapter was just the setup for Jodies later break with the system. If you look at it as a gaming element then it was a bit weak. But it was a plot device and as that it set the scene perfectly.

 

[Sigfried1977]

I know what its purpose was (it's not like Cage is a man who's ever been in danger of so much as flirting with a concept like ambiguity). Doesn't make it any less lousy as a piece of story. The ending in particular was just terrible. The whole thing was an exceptionally clumsy plot device, and if Cage wasn't so in love with Hollywood's fetish for gasoline explosions, he could have easily implemented something plausible instead.

Spoiler

she's working for the CIA for Christ's sake. Shouldn't have been too hard to come up with a believable way for her employer to screw her over. But no, the game wants me to believe that a CIA operative - a CIA operative who's worked on foreign soil before no less - is somehow oblivious to the fact that she's being sent out to assassinate "the democratically elected president of Somalia"

I'm fairly good at suspending my disbelief, but when the Somalia chapter came to its bafflingly stupid conclusion, my suspension cable snapped.

 

[Laa-Yosh]

What this game demonstrates visually is that the artists working in games today can be as proficient as the ones working in movie VFX and the tech is less and less of a bottleneck. Also, the approahces are pretty much the same, the main difference is that games have to work with more or less crude approximations instead of the 'real' thing. But the principles are as close as it gets, compared to what was going on 10 or even 5 years ago.

 

[patsu]

What is the turnaround time like for changes, say if someone wants to change the main character's look ?

 

[ultragpu]

What this game demonstrates visually is that the artists working in games today can be as proficient as the ones working in movie VFX and the tech is less and less of a bottleneck. Also, the approahces are pretty much the same, the main difference is that games have to work with more or less crude approximations instead of the 'real' thing. But the principles are as close as it gets, compared to what was going on 10 or even 5 years ago.

I heard even offline movies cheat all the time, such as GI approximation for Avatar and SSS hack in Prometheus. The real thing according to them were just too time consuming, so perhaps one could say movies are highly glorified versions of real time games.
And of course they may be some exceptions which I can't pin point myself.

 

[Laa-Yosh]

Your concept of what's considered cheating in VFX is wrong. The difference from games is that image quality can not be compromised.

Yeah, Weta has a new SSS implementation that may render faster; but the main point is that it's also higher quality compared to the previous iteration. Also, their Pantaray renderer for baking GI lighting into spherical harmonics is not really a cheat, just a different approach (and still quite resource hungry).

The other approach to "cheating" is to simply eliminate unnecessary work - use matte paintings or 2.5D backgrounds whenever the camera movement allows it, fix small image problems in post manually, fix 3D deformation problems on cached vertex data instead of the character rig, and so on. Again, the guiding principle is to keep the expected level of quality but use the most resource efficient method available.


Games on the other hand will compromise image quality. This is a major difference and I thought I was quite clear in outlining the details already.

 

[Laa-Yosh]

What is the turnaround time like for changes, say if someone wants to change the main character's look ?

That depends on the extent of changes... Small tweaks won't have major consequences, but others may require throwing away textures, rigging data, possibly even the existing mocap (QD's approach is based on 100% similarity between the CG character and the p-cap talent).

If you can define it more precisely, I can get into more detail...

 

[patsu]

That depends on the extent of changes... Small tweaks won't have major consequences, but others may require throwing away textures, rigging data, possibly even the existing mocap (QD's approach is based on 100% similarity between the CG character and the p-cap talent).

If you can define it more precisely, I can get into more detail...

The worst case is to throw the entire character away (3 months' worth of work ? I have zero clue)

How 'bout just a handful of common cases.

Guerilla is b*tching about nextgen asset costs. So beyond the easier technical development, just how big a cost is assets now ? In the end, nextgen development cost still ballooned right ?

 

[Laa-Yosh]

Well, if I had to add up all the work we do on a full character and its rig and facial blendshapes, that 3 month guess wouldn't be too far off. Most games might be able to get away with a little less... but it also depends on whether you'll count time spent on gameplay animation cycles (which we don't need to do). I do recall Epic throwing out 6-8 weeks just for the asset creation, and rigging can be time consuming as well (Gears usually didn't require much because of the armor design, but once you start to have cloth and hair...)

Then again this is for the most detailed and polished "hero" assets. Those are usually 10-30% of all the characters made for a project, and I guess it applies to games just as well as to our work (CG). Background guys at us are built on a much more streamlined pipeline, making the most of re-using stuff and simplifying both the asset and the workflow (fewer hands touching it means less coordination and less time).

For the costs, just start with the average game dev yearly salary and divide it by something between 4 and 12. Usually even in studios with generalists (like Blur) there's at least 2-3 different people involved (possibilites are: concept artist, modeler, sculptor, texture artist, rigging guy, shading guy, hair guy...) but it's easier to add up all the time and multiply it by the average salary.

You also have to consider costs for hardware, software, office space and such; then the cost of either building or renting facilities for motion/performance capture and scanning (these can be mighty expensive, even if you can amortize the costs over multiple characters and projects).
Also, hiring actors for mocap, and/or likeness and facial expression capture and such, can be quite expensive too. Real life talent may also require legal backup (so that they can't sue the publisher later) and lawyers aren't cheap either, even if you rely on modeling agencies and such.

So this stuff is getting really complex nowadays. Oh and another example, we're now consulting with a real life costume designer on our current project to improve our cloth stuff, both the models and the simulations. She's worked on a lot of stuff from theater and opera to TV series like the Borgias, and it's a really interesting experience, finding out how and why costumes are constructed, drawing patterns, learning the history of it and all. We're probably going to hire someone from that industry to work on site soon, too. I expect that sooner or later games will have to do this too.


Common cases... actually there aren't many in our case as our schedules are pretty compressed, just a few months instead of years. Not sure how it goes with games, as they tend to spend much more time on tweaking faces and deformations and such for leads because that stuff is seen most of the time, especially in TPS games.

However we sometimes get client feedback on likeness, like it looks too much like the celebrity they chose for inspiration, or it looks different once final cameras and lighting are used. This is relatively easy to fix, add a corrective blendshape and fix anything that gets messed up in turn. I guess games have to rebuild the asset instead as they usually can't cheat like this (memory issues and such).
There are many other similar possibilities, like finding out in practice that you've messed up some of the proportions or so, these are also relatively easy and cheap to fix.
More costly changes would involve a radically different facial structure, a redesign of the costumes, and so on. Sometimes you can salvage existing stuff, but in general you'd need time proportional to the amount of geo/texture you replace.

 

[patsu]

Yes, I am most interested in animation changes (face and body), because I think it's inherently tied to AI and gameplay.

If asset demands the most work nextgen, it sounds like your company will be rich soon.

 

[Laa-Yosh]

Animation changes are far cheaper in games, as nothing has to be pre-calculated. There's a small overhead for manually polishing cutscene-specific unique animations but it's probably not too significant.

New Cryengine games might be different though, as they can bake any deformations from Maya into Alembic format and stream it during cutscenes. This may include expensive simulations for cloth, hair, even muscles - the realtime cost remains the same as the game just has to load vertex transformation data.


As for our company, we're still not involved in asset production although from time to time, some of our stuff ends up in game content (used as highres source meshes for the ingame models' normal and color maps). For now, we're still booked doing CGI work in the foreseeable future

 

[patsu]

Animation changes are far cheaper in games, as nothing has to be pre-calculated. There's a small overhead for manually polishing cutscene-specific unique animations but it's probably not too significant.

Can you mix mo-capped and computer generated animation freely now ? Like making B2S's mo-capped characters behave differently using computation ? The computer-generated movement and the mo-capped movement may clash in style but is it feasible today in the first place ?

New Cryengine games might be different though, as they can bake any deformations from Maya into Alembic format and stream it during cutscenes. This may include expensive simulations for cloth, hair, even muscles - the realtime cost remains the same as the game just has to load vertex transformation data.

I always assumed this was something the game industry is already doing.

As for our company, we're still not involved in asset production although from time to time, some of our stuff ends up in game content (used as highres source meshes for the ingame models' normal and color maps). For now, we're still booked doing CGI work in the foreseeable future

Ah, I always thought sooner or later, you'll be involved in game asset production. The line is blurring right ?

 

[Laa-Yosh]

Can you mix mo-capped and computer generated animation freely now ?

I think games like GTA4 were already doing smooth transitions from mocap/keyframe into Euphoria procedural animation for years now. Also, Crysis 1 was blending anim cycles and realtime inverse kinematics... in fact even Halo1 had IK to plant feet on the ground AFAIK.

Like making B2S's mo-capped characters behave differently using computation ? The computer-generated movement and the mo-capped movement may clash in style but is it feasible today in the first place ?

Computer generated movement has its limits even today. You cannot make someone walk or emote or do anything expressive because computers are dumb and have no artistic sense. This is still researched and I haven't seen anything even remotely convincing.
In fact even keyframe animation - stuff created from scratch by an animator - is usually better for cartoon style characters, any kind of realistic human motion has to be motion captured or rotoscoped using video reference.

You can however make someone fall, or modify the existing animations to follow a target with the head's rotation or plant the feet or blend between two animations and so on.
You can also do automatic lip sync based on audio files, or mix together different anim clips using scripting like the Mass Effect games did.

I always assumed this was something the game industry is already doing.

Yeah, but only to a certain extent; it's mostly about modifying existing anim cycles for dynamic situations by using inverse kinematics or various constraints (aim, orient etc) - or generating physics-based reactions to ingame events (ragdoll stuff mostly).

Ah, I always thought sooner or later, you'll be involved in game asset production. The line is blurring right ?

Well, I'm not sure. I've been seeing actual nextgen assets up close for a while now and we're still far ahead of those specific games... The Dark Sorcerer stuff or Ryse is a bit different, they're closer to what we do, so these projects could hire us for outsourcing.

However the asset team is still just one part of our staff and we have a lot of other talent - riggers, animators, compositors, lighting and shading guys, the director and his team who create the cinematic stuff... And of course there's the production and coordination team and backoffice and mocap and stunt guys and so on
So we're still in the film business mostly, and our recent projects all had enough asset work to keep us busy enough.

 

[Laa-Yosh]

Oh and those nextgen games, they all have reasons why the assets aren't like what you can see in Ryse...