It's this sort of thing that gets my blood boiling oh so wonderfully for gaming again. For a while there in the recent past I'd kinda been losing my yearning for it, seeing the same old stuff and what not. But then things like this pop up every so often...
Thanks for posting it, wouldn't have caught it otherwise.
Great short, but tech wise its a bit disappointing. It's lacking in facial animations, as well as it's not true motion scanning, a lot of clean up work done on the textures and models.
Coooool! I'm going to rain on some parades and say the facial animation was disappointing, particular the lip syncing. It's like she could never fully close her mouth.
Loved the premise and the acting. Definitely tugged at the heart strings a bit.
However, the facial animation is lacking. Her lips don't always touch when they need to, to pronounce a word. Her lips don't purse shut at all it seems. It's still a bit robotic, it needs more pop. Eyes looked good though.
It's interesting they went for the easiest haircut they could find, and left it out for most of the video. For all the blockiness the graphics of 15 years ago had, they at least could make curly hair.
Also, a question. Isn't it possible to make a special texture to darken teeth that depends on the animation, like a wrinkle map but with a diffuse texture? It's becoming more obvious through the years that teeth just can't be shaded properly by traditional methods, and they stand out badly in this video.
If Kara didn't last 5 seconds after being assembled before she thought something, then in the game world she's going to get into trouble shortly after the end of the first chapter. The rest will be an Asimov / Blade Runner story-based game (similar to Heavy Rain - nothing wrong with that) with sections dealing with humanity, psychological troubles likely including Stockholm Syndrome, as well as a range of underworld bot dealers. All may or may not be resolved if/when the player makes it to the ending.
I don't see why they wouldn't continue with this. Unless there's a fantastic game Quantic are making instead, this has to be up there with some of the better concepts and themes for a game.
Funny how so much effort went into making Kara feel lifelike and human, yet the actual human's line delivery was really artificial and stiff. Not knocking it or anything, this was a great demo and I love this sort of robot-rights sci-fi, I just found it ironic.
Fingus, it's a guy talking to what he believes is a "thing" - no need for him to have emotion at all in what he says.
The short was great, but the mouth animation left a lot to be desired. I recall even HL2's to be better and it's definitely not up to spec with LA Noire animation. The eyes were really well done though. I'd love to see a game made of it.
Fingus, it's a guy talking to what he believes is a "thing" - no need for him to have emotion at all in what he says.
The short was great, but the mouth animation left a lot to be desired. I recall even HL2's to be better and it's definitely not up to spec with LA Noire animation. The eyes were really well done though. I'd love to see a game made of it.
The eyes are what got me. I thought I was just watching a tech demo at first... and then I couldn't pause it. Holy shit.
this demo, just like their previous demo all have pretty awkward lip animations. Something about the corners of her mouth when she smiles and the whole cheek area not stretching as it should kinda ruins the effect. The tech isn't there yet, but its getting closer.
The eyes are pretty convincing although I miss the clear wrinkle above her eyes
Lacking in facial animation? I thought it looked pretty good. I totally related to the character and had no problems buying into the premise.
Well, holding LA Noire and some of the motion scanning tech I've personally worked with as the standard, the lips and eyebrows aren't true motion scanning. It's probably mostly hand animated, with maybe some of the motion scanning used as reference. The model also looks like it had some work done to it as well.
It's a fantastic demo none the less, just not what it advertised at GDC.
I love stuff like this. And hopefully it leads to some interesting gameplay in the future. I feel games today are far to reliant on combat for everything.
Well, holding LA Noire and some of the motion scanning tech I've personally worked with as the standard, the lips and eyebrows aren't true motion scanning. It's probably mostly hand animated, with maybe some of the motion scanning used as reference. The model also looks like it had some work done to it as well.
I don't think there's any "motion scanning", not the way it was done in LA Noire.
I think it's a single face camera and 2D image analysis, maybe with Image Metrics software, maybe with something custom. Then the data drives some carefully hand crafted expressions on a joints based rig.
What's missing is the facial folds and wrinkles, but that shouldn't be that hard to implement, Uncharted 1 used it four years ago already. I wonder how realistic they want their fictional android to be - it may also need sticky lips and the eyelashes could need something other than textured poly strips.
I don't think there's any "motion scanning", not the way it was done in LA Noire.
I think it's a single face camera and 2D image analysis, maybe with Image Metrics software, maybe with something custom. Then the data drives some carefully hand crafted expressions on a joints based rig.
What's missing is the facial folds and wrinkles, but that shouldn't be that hard to implement, Uncharted 1 used it four years ago already. I wonder how realistic they want their fictional android to be - it may also need sticky lips and the eyelashes could need something other than textured poly strips.
Yea, it looks like the Image Metrics stuff I've seen a lot in the past, but they were going around GDC yesterday saying it was motion scanning so that's why I was all Marge Simpson face.
I was at the GDC talk. The tech demo advocates full performance capture, where the body, face and voice are all recorded simultaneously. I think there were 90 markers on her body, and 90 on her face, and apparently a sound proof recording area was a difficulty. And all that, somehow, was supposed to be the reason behind the emotional response we feel.
It's a great tech demo. I don't deny that. I just feel kind of "marketed" at, in the context of the presentation at GDC. To me, the emotion isn't created by full performance capture. It's created by having a solid script and a great actress, with an amazing voice. I honestly believe you could have a blank screen, and still get chills hearing the performance. It might even be better that way, as you wouldn't be looking at how strange her mouth looks.
A guy asked how he, as an Indie Dev, could make use of the info in the talk. The presenter said you just need to be able to make that kind of investment if you want this kind of emotion. Wrong?
really if this was a full game with a really immersive story, i would say the tech is good enough, and there is nothing there that is big enough to pull me out of that immersion.
just getting torn apart now since it is just a short clip, with no additional story, context or attachment to characters you would get from a full game.
Yeah as Elyaradine said, it's full motion capture, everything is pretty much picked up in one go.
I'll have to post up some of the work I did with a research group at USC. It's full motion capture like this too. We had markers on the face and body that captured obviously the body and face animations/models but also got diffuse, normals, specular, and subsurface scattering diffuse. The lips should be more accurate if it's the same process I'm thinking of. But oh well.
Well, holding LA Noire and some of the motion scanning tech I've personally worked with as the standard, the lips and eyebrows aren't true motion scanning. It's probably mostly hand animated, with maybe some of the motion scanning used as reference. The model also looks like it had some work done to it as well.
It's a fantastic demo none the less, just not what it advertised at GDC.
Yeah after going back and comparing to LA Noire I can totally see the difference and see what your talking about. I'd love to see some of that USC work you mentioned earlier.
Found this on Reddit, and it puts a whole new perspective on the video. I quite enjoy this.
Everything the man said to the android was to confirm that it COULD feel emotion. The condescending and dismissive tone. The casual finality in his voice about being sold as merchandise. Even his faux reaction to her sentience. If she had simply said "yes, I understand" when confronted with being disassembled, then they actually would have had to disassemble her. It's because she did show emotion that she was confirmed to be functioning proper.
They are making emotion capable androids.
Everything the man said to the android was to confirm that it COULD feel emotion. The condescending and dismissive tone. The casual finality in his voice about being sold as merchandise. Even his faux reaction to her sentience. If she had simply said "yes, I understand" when confronted with being disassembled, then they actually would have had to disassemble her. It's because she did show emotion that she was confirmed to be functioning proper.
They are making emotion capable androids.
That's a REALLY interesting way of looking at it. Some of the supervisor's responses at the end throw it off a little, but does put the entire sequence in a very different frame.
To me, the emotion isn't created by full performance capture. It's created by having a solid script and a great actress, with an amazing voice. I honestly believe you could have a blank screen, and still get chills hearing the performance. It might even be better that way, as you wouldn't be looking at how strange her mouth looks.
I dunno, to me the visuals were pretty crucial. Especially the eyes - there was an amazing amount of pathos there, and they really carried a huge amount of the main character's emotion, which was the linchpin of the entire sequence.
i think it was very well balanced along all the elements. Ofcourse some things could have looked better but overal this is one of the better realtime cinematics i've seen especially if you consider it's running on a ps3 (or so i think)
anyway i believed the message they were bringing across and that's what counts in the end
I'd say, controller guy fucking suck at his job, and should be fired asap
smash those emotional robots to shit, because we all know how it ends
[ame="http://www.youtube.com/watch?v=BAjdlwnTg5w"]Animatrix the Second Renaissance Part 1 - YouTube[/ame]
May I ask if the face is using 3D markers or a single face cam and 2D image analysis?
No face cam was used, so 3d markers. The recording was done in one take, with two actors mimicing the actions of the robotic arms. There was a great number of cameras (don't remember the quoted number) recording the face and body at the same time. He said that he wanted to avoid any sort of face cam because it makes any sort of close facial interaction between actors difficult.
Furthermore, that was an old version of their render engine (A year if I remember correctly) running in real time on a PS3. There was also minimal clean up so he said, and the goal of the clean up was simply to retain the integrity of the original actress' performance.
Anyway, I'm not sure I'm really into this sort of a set up. I've gotta question the practicality of using casting actors for games when you can't know what a player will do. Seems great for cutscenes, but I don't know what you could really do beyond that. It'll be fun to watch I guess.
This was pretty impressive as a real-time tech demo, but I get the feeling that David Cage is just a frustrated film maker who's stuck making tech demos for Sony. I really enjoyed the early parts of Fahrenheit, with its branching narratives, but I thought Heavy Rain was really quite disappointing. The interactive parts of the game were quite limited and the story was solid enough for a game, but when compared to films it came off as a derivative serial killer potboiler. I hope that Quantic Dream's ambition for their next game stretches to game design too. They seem keen to keep pushing the envelope with technology that creates a barely interactive experience that's attempting to mimic, but is actually a lot less believable and immersive than a movie.
I wasn't very clear on what I could say or not (not part of the animation team), but I guess most of it has been explained at the presentations as what Daelus describes is spot on. Full body + face 3D markers, no extra cams picking up any kind image info (would be curious to see any of the stuff you did at USC BlvdNights).
Characters are made with the usual Zbrush + Maya + Photoshop, with a complex facial (and body of course) rig to receive the animations.
This was pretty impressive as a real-time tech demo, but I get the feeling that David Cage is just a frustrated film maker who's stuck making tech demos for Sony.
my thoughts exactly.
generally speaking - shoehorning movie-style ways of storytelling into games doesn't work for me at all. cutscenes must die. it's an interactive medium, last thing i want is to be forced to take a break and watch my avatars do something the interactive bits won't let me achieve myself.
Replies
Thanks for posting it, wouldn't have caught it otherwise.
However, the facial animation is lacking. Her lips don't always touch when they need to, to pronounce a word. Her lips don't purse shut at all it seems. It's still a bit robotic, it needs more pop. Eyes looked good though.
But yeah, the lip syncing was quite iffy.
But that was some pretty damn fine work, when you look at everything together. Amazing story telling, music, just fantastic.
Of course knowing the devs, this will all be for nothing when the scene after it is simply "Press X fast to fuck the real doll".
Also, a question. Isn't it possible to make a special texture to darken teeth that depends on the animation, like a wrinkle map but with a diffuse texture? It's becoming more obvious through the years that teeth just can't be shaded properly by traditional methods, and they stand out badly in this video.
If Kara didn't last 5 seconds after being assembled before she thought something, then in the game world she's going to get into trouble shortly after the end of the first chapter. The rest will be an Asimov / Blade Runner story-based game (similar to Heavy Rain - nothing wrong with that) with sections dealing with humanity, psychological troubles likely including Stockholm Syndrome, as well as a range of underworld bot dealers. All may or may not be resolved if/when the player makes it to the ending.
(purely speculation)
Zwebbie, yeah the teeth are less of a problem with proper subtle SSAO nowadays I think.
The vid is already more than a year old, a bit unfortunate it wasn't shown earlier. I made the pulsating heart thingy!
I don't see why they wouldn't continue with this. Unless there's a fantastic game Quantic are making instead, this has to be up there with some of the better concepts and themes for a game.
simps... bjork did it
The short was great, but the mouth animation left a lot to be desired. I recall even HL2's to be better and it's definitely not up to spec with LA Noire animation. The eyes were really well done though. I'd love to see a game made of it.
The eyes are what got me. I thought I was just watching a tech demo at first... and then I couldn't pause it. Holy shit.
The eyes are pretty convincing although I miss the clear wrinkle above her eyes
Well, holding LA Noire and some of the motion scanning tech I've personally worked with as the standard, the lips and eyebrows aren't true motion scanning. It's probably mostly hand animated, with maybe some of the motion scanning used as reference. The model also looks like it had some work done to it as well.
It's a fantastic demo none the less, just not what it advertised at GDC.
I don't think there's any "motion scanning", not the way it was done in LA Noire.
I think it's a single face camera and 2D image analysis, maybe with Image Metrics software, maybe with something custom. Then the data drives some carefully hand crafted expressions on a joints based rig.
What's missing is the facial folds and wrinkles, but that shouldn't be that hard to implement, Uncharted 1 used it four years ago already. I wonder how realistic they want their fictional android to be - it may also need sticky lips and the eyelashes could need something other than textured poly strips.
Yea, it looks like the Image Metrics stuff I've seen a lot in the past, but they were going around GDC yesterday saying it was motion scanning so that's why I was all Marge Simpson face.
It's a great tech demo. I don't deny that. I just feel kind of "marketed" at, in the context of the presentation at GDC. To me, the emotion isn't created by full performance capture. It's created by having a solid script and a great actress, with an amazing voice. I honestly believe you could have a blank screen, and still get chills hearing the performance. It might even be better that way, as you wouldn't be looking at how strange her mouth looks.
A guy asked how he, as an Indie Dev, could make use of the info in the talk. The presenter said you just need to be able to make that kind of investment if you want this kind of emotion. Wrong?
On an unrelated note, sex doll. mmm.
just getting torn apart now since it is just a short clip, with no additional story, context or attachment to characters you would get from a full game.
I'll have to post up some of the work I did with a research group at USC. It's full motion capture like this too. We had markers on the face and body that captured obviously the body and face animations/models but also got diffuse, normals, specular, and subsurface scattering diffuse. The lips should be more accurate if it's the same process I'm thinking of. But oh well.
May I ask if the face is using 3D markers or a single face cam and 2D image analysis?
Instead... "holy shit this andriod is crying and pleading for her life... I've never seen that before. Oh well, i'll box her up with the rest"
looks cool tho
Edit:
Also, they should make this thing as the frontend to the PS3. Like the Sony version of the Xbox avatar.
Yeah after going back and comparing to LA Noire I can totally see the difference and see what your talking about. I'd love to see some of that USC work you mentioned earlier.
Mind you, I get the feeling that Quadratic dream is becoming the "Girl crying in our trailer" company.
I dunno, to me the visuals were pretty crucial. Especially the eyes - there was an amazing amount of pathos there, and they really carried a huge amount of the main character's emotion, which was the linchpin of the entire sequence.
and why would you want to create a product that does not want to be sold?
anyway i believed the message they were bringing across and that's what counts in the end
smash those emotional robots to shit, because we all know how it ends
[ame="http://www.youtube.com/watch?v=BAjdlwnTg5w"]Animatrix the Second Renaissance Part 1 - YouTube[/ame]
I thought it was going to more this way when i was 3 more of them ::]
[ame="http://www.youtube.com/watch?v=d2tBhaVEWGM"]Björk - All is Full of Love (HD 720P) - YouTube[/ame]
No face cam was used, so 3d markers. The recording was done in one take, with two actors mimicing the actions of the robotic arms. There was a great number of cameras (don't remember the quoted number) recording the face and body at the same time. He said that he wanted to avoid any sort of face cam because it makes any sort of close facial interaction between actors difficult.
Furthermore, that was an old version of their render engine (A year if I remember correctly) running in real time on a PS3. There was also minimal clean up so he said, and the goal of the clean up was simply to retain the integrity of the original actress' performance.
Anyway, I'm not sure I'm really into this sort of a set up. I've gotta question the practicality of using casting actors for games when you can't know what a player will do. Seems great for cutscenes, but I don't know what you could really do beyond that. It'll be fun to watch I guess.
Characters are made with the usual Zbrush + Maya + Photoshop, with a complex facial (and body of course) rig to receive the animations.
recent swedish scifi-drama real humans:
[ame="http://www.youtube.com/watch?v=VKOZoAdhHso"]SVT: Real humans - YouTube[/ame]
my thoughts exactly.
generally speaking - shoehorning movie-style ways of storytelling into games doesn't work for me at all. cutscenes must die. it's an interactive medium, last thing i want is to be forced to take a break and watch my avatars do something the interactive bits won't let me achieve myself.
Quite impressive then, wonder about their systems...