Uncharted 3
- Thread starter corduroygt
- Start date
No, that is totally different 
Did you see any actors wearing blue make-up in the "making of" videos? No you didn't... There are three basic methods to capture an actor's facial performance.
MotionScan basically records everything using a lot of normal RGB camera inputs. They build a mesh, a normal map and a color texture, for every frame, and it contains all visible information - lighting, reflections, hair, make-up. Then they put this animated mesh + texture combo on top of an animated but headless character. Very limited, can't animate it, can't replace the looks of the actor, but also highly realistic, although it also depends on the mesh + texture resolution. PS4/Xbox3 will be able to produce even better and more realistic results.
This is the newest tech as far as I can tell, too.
With traditional facial motion capture they use optical markers and a lot of Vicon's infrared cameras. They track each marker and return their XYZ positions and use that to more or less directly deform the face geometry. This usually results in bad deformations and zombie characters like Polar Express or Beowulf. This isn't an entirely realistic method as it can't handle even small differences between the real actor's face and that of the CG character, it can't handle complex information like eye rotations, interior side of the lips and mouth, and so on. It's also the oldest method but it's been developed further and further in the past decade, nevertheless, it's probably going to be left behind soon as it's not good enough.
Image analysis based performance capture is different. It tracks the actor's face and tries to analyze it to determine which basic facial movements are used to build the expressions. Stuff like the intensity of blinks, smiles, jaw movement, brow wrinkling and so on. This generates a set of metadata that's used to drive a more complex facial rig that has proper deformations assigned to these elementary facial movements. The level of realism depends on the artists and tech people who've built the facial deformations, and it usually also requires animator input (which also means that there is proper animation and it can be modified). It's the most advanced method, it can also be paired with full body motion capture to get the entire performance, and it's also subject to a lot of more research.
This is also how Avatar's system works, image analysis and a LOT of manual, traditional keyframe animation used on a very very refined, hand crafted deformation rig.
Did you see any actors wearing blue make-up in the "making of" videos? No you didn't... There are three basic methods to capture an actor's facial performance.MotionScan basically records everything using a lot of normal RGB camera inputs. They build a mesh, a normal map and a color texture, for every frame, and it contains all visible information - lighting, reflections, hair, make-up. Then they put this animated mesh + texture combo on top of an animated but headless character. Very limited, can't animate it, can't replace the looks of the actor, but also highly realistic, although it also depends on the mesh + texture resolution. PS4/Xbox3 will be able to produce even better and more realistic results.
This is the newest tech as far as I can tell, too.
With traditional facial motion capture they use optical markers and a lot of Vicon's infrared cameras. They track each marker and return their XYZ positions and use that to more or less directly deform the face geometry. This usually results in bad deformations and zombie characters like Polar Express or Beowulf. This isn't an entirely realistic method as it can't handle even small differences between the real actor's face and that of the CG character, it can't handle complex information like eye rotations, interior side of the lips and mouth, and so on. It's also the oldest method but it's been developed further and further in the past decade, nevertheless, it's probably going to be left behind soon as it's not good enough.
Image analysis based performance capture is different. It tracks the actor's face and tries to analyze it to determine which basic facial movements are used to build the expressions. Stuff like the intensity of blinks, smiles, jaw movement, brow wrinkling and so on. This generates a set of metadata that's used to drive a more complex facial rig that has proper deformations assigned to these elementary facial movements. The level of realism depends on the artists and tech people who've built the facial deformations, and it usually also requires animator input (which also means that there is proper animation and it can be modified). It's the most advanced method, it can also be paired with full body motion capture to get the entire performance, and it's also subject to a lot of more research.
This is also how Avatar's system works, image analysis and a LOT of manual, traditional keyframe animation used on a very very refined, hand crafted deformation rig.
Also, to clarify something else, an animation rig is basically a character (or face) with a set of controls to change its shape - open the mouth, put on a smile, blink etc. You can combine these elemental expressions into anything you want.
With motionscan, you have nothing. If the actor choses to blink at a certain point of time, it's part of the entire model and the texture, there's no eyeball or eyelids, you can't move in and change it at all.
With motionscan, you have nothing. If the actor choses to blink at a certain point of time, it's part of the entire model and the texture, there's no eyeball or eyelids, you can't move in and change it at all.
I can't believe I just noticed that when Drake goes inside the ship you can see the chandeliers moving side-to-side.
Oh, and when he goes into an office and a bottle is rolling from one side to another.
I've seen the demo plenty of times, but it's amazing that I'm just noticing those small details just now.:smile:
Naughty Dog isn't capturing any facial performance, they only film it with HD cameras as reference but the animation is 100% keyframed. It's more of an artistic choice.
Most other games today use a combination of image analysis based perf capture and keyframe animation if there's a lot of content. AC, Crysis, GTA and so on are some better known examples.
Most other games today use a combination of image analysis based perf capture and keyframe animation if there's a lot of content. AC, Crysis, GTA and so on are some better known examples.
L. Scofield
Veteran
You mean the projected shadow?
UC2 used 2xMSAA that was half broken but still looked okay much of the time. UC3 will use some post-process AA iirc.
Billy Idol
Legend
imo AA is super important...shimmering, aliasing and crawling edges destract me the most...get a clean image first and then look for the rest 
that said...I hope ND improves their AA solution in UC3!
the action on the other hand in this demo looked great..."want to play this now"-great
that said...I hope ND improves their AA solution in UC3!
the action on the other hand in this demo looked great..."want to play this now"-great
I also have mixed feeling about the demo. The water tech was really nice and I'm really interested in how it works underneath, and the lighting was more impressive then U2. But it seemed to me like both had their shortcoming: water flow when the ship breaches seemed crude - like it's just not flowing right (perhaps we are just not there yet?), and as lighting goes it seemed that some light sources (mostly chandeliers and spotlights) look better then others. But it's hard to judge properly.
It didn't help that the level itself seemed a bit boring when trying to lean too much towards Splinter-Cell stealthiness in closed environments.
Besides, what's up with everyone showcasing ship levels this E3? It didn't do Uncharted grace showing up right after we saw the Tomb Raider ship pre-rendered trailer a few days earlier - a harsh remainder that games are still far away from CGI quality. And then we had a Modern Warfare submarine stage demo, a Gears of War ship boss stage demo, and an Uncharted 3 ship stage demo (oh, and throw in a wrecked spaceship teaser for Halo 4).
I don't know what was more dissapointing: when I saw Gears demo and thought to myself "hey, isn't this the Uroboros level from Resident Evil 5?" or when I saw the Uncharted demo and thought "mmm, isn't this the first Cold Fear level?" You know - all the way down to the stormy whether, water on deck, rocking ship, balancing animations etc.
Overall, the Uncharted 2 E3 stage demo looks much more impressive to me even after watching the Uncharted 3 demo so it's a bit of a dissapointment. At least the new trailer looks great.
I understood need to respect the personal opinion but really what have of less impressive the last gameplay of U3?
The scale is a lot more huge than U2 in every aspect, procedural water shader are really astonishing...Everything was fine besides the sounds of the guns. Their last E3 demo was cool, but the gameplay in that segment was limited compared to this one. And the moment to moment transitions were very good throughout once the water started pouring in. Didn't bother me one bit that it ended on a cliffhanger.
I am really blown away by the lighting (the one highlighting the stealth action against the first guard), Drake's animations (the best I've seen in a videogame as of yet...) and the whole attention to detail (water shaders, water simulation in the pool and the raging sea, overall direction...). But truth be told, some of the face close ups were quite horrible (like when Nathan resurfaces after diving to breath...). What's the matter with that? :smile:
I love the beefed up stealth elements because the previous Uncharted games have too much shooting. Here you can basically choose your own style.
Use of water is great but it is the combined use of lighting, water, animation/organic combat and tumbling space with movable/swaying objects altogether that makes this trailer stand out for gamers compared to just cutscenes. As you can tell from Heavy Rain and LA Noire threads, some gamers don't like watching cutscenes or so-called movie games.
Use of water is great but it is the combined use of lighting, water, animation/organic combat and tumbling space with movable/swaying objects altogether that makes this trailer stand out for gamers compared to just cutscenes. As you can tell from Heavy Rain and LA Noire threads, some gamers don't like watching cutscenes or so-called movie games.
Prophecy2k
Veteran
So does that mean that the actors whilst doing their performances had to be extra careful not to even blink??
whoooaa!I wonder how it would look if they sneezed
lolGoing back to a comment you made earlier about their inability to capture eye movement, it's funny because whilst playing through LA Noire, after reading your comments, i was always subconciously watching the eyes to see when they'd slip up and it would look really odd. I did it so much i ended up failing loads of my interviews for not paying proper attention to what the characters were actually saying (thanks Laa-Yosh)
I did think they did a good job with the eyes in LA noire though. And whilst i enjoyed the motionscan tech in that game, i don't think i'd want it Uncharted 3. Even if drake's voice actor actually looked like drake. I think the super realistic emotions and expressions would take away alot of Drake's charm for me. Uncharted was never a series that aspires to super realism in it's art and character designs imho. And it's all the better for it
I do wonder whether GTAV will use motionscan though...
Avatar's sytem works with just make-up, no need for reflective markers. I think they can also track eyes to an extent, using image analysis. And yeah, 90% of the performance is from the facecam capture. Then again it's usually the last 5% that's the hardest and most work intensive to get right.
