Slashdot Mirror
More Realistic Rendered Flesh
The Renderman writes "Check out this ubercool new rendering technique for skin from the guys who invented photon mapping. They also have several animations to demonstrate the technique - all rendered using Linux :) It makes those faces in Final Fantasy look like plastic. This technology should make the actors in Hollywood
more nervous than what they saw in FF." The examples prove that human skin needs more blemishes. Without zits, pimples, moles, or scars, the skin still looks fake, it is definitely a cool step forward.
Sigh. Please read the explanation here.
One thing that I don't see anyone giving a solution to it - are the eyes - you play around the eyes, adding wrinkles, try to make the image look happy, sad, angry - but the eyes themselves looks like a dead person's eyes...
Hetz (Heunique)
Amen.
There are, of course, tons of examples of existing movies where there's a big difference between the what appears on the screen and the actor behind the character: animation, the same actor playing multiple characters, different actors doing the voice and the body for the same character, etc.
On the other hand, there's still the major problem of creating affect: it's very difficult to draw or paint or model a person displaying a certain emotion without working from a real person, and doing the voice is at least as hard (besides the fact that speech synthesis is not as good as rendering yet).
I suspect that, even for an entirely CG movie, the studio would want human actors for all of the parts, so that the animators could see how each character's face looks when they are doing what the script says.
Furthermore, there's a lot more to each character than what's in the script; the actor (and director) determine what the character is thinking and feeling while doing the predetermined actions. It's essentially the Turing Test: convince an audience that you really are a person in a certain situation. Neither a computer, nor even a group of graphics researchers are going to be able to get into the roles that way any time soon.
This could, of course, mean that an actor's appearence makes very little difference; a skillful actor could play a character with no physical resemblence if the task involved only showing a graphics team what the necessary emotions look like.
The hardware manufacturers need to have a reason to do this algorithim (or do the compute-expensive portion of it) and (if it is also technically possible) you will see it.
Almost everything the cards does is algoirthims invented 30 years ago. The clever engineering is getting them to run fast and to correctly select a subset of the possible algorithims to implement. There are also hundreds of graphics algorithims invented 30 years ago that are *not* on any hardware card.
If somebody invents a new algorithim that uses various derivatives of the normals and surface position to produce some new lighting effect, they can be implemented in the pixel shaders. But if the algorithim requires volumetric information that is an integral of all the surrounding surface points, the information is not there, because the hardware is not producing it!
Yea, you can use some pixel shaders to index a 3-D volume that may be pregenerated by this algorithim, and this is "faster". But that is equivalent to pre-rendered radiosity being loaded into texture maps, most people do not consider that "hardware radiosity", and I don't think this pixel shader solution would be considered "hardware photon mapping".
I love it when uninformed idiots spout shit like this. "Ooh! I haven't seen the movie yet, but they're CG people! They have to look plastic! All CG people look fake."
Have a look at this link, and pick through the process of creating the characters. Then try to say they look like plastic.
I applaud the work of the Final Fantasy CG people. All of them. Their characters are the most realistic performers (outside of real actors) to ever hit the screen.
-- Give him Head? Be a Beacon?
-- Give him Head? Be a Beacon? :P)
(If you can't figure out how to E-Mail me, Don't.
The *concept* of CG actors may leave people cold on the face of it. But the reality of seeing those digital actors in an actual context... engage the viewer in a compelling story, and he or she will not be paying attention to whether the actor is digital, flesh and bone, or hand drawn.
I bet if you asked these same people about using technology to revive, say, John F. Kennedy from the grave to appear in a movie, they would have also been dubious. But that didn't stop Forrest Gump from being a success, mostly because of that technology. In the context of the film, it made sense and looked fairly natural.
Where digital actors will make an impact first (and have already been making an impact) are background crowd scenes and walk-ons. With some motion capture equipment, a few actors can become a huge crowd that does exactly what the director tells them, every time. The A-list stars may not be threatened, but I suspect people who make some money on the side by being extras will be seeing less work. I also suspect stunt people will be seeing less work as well. Why risk a real human life by having them jump off a building, when you can get a digital actor to jump and come to a splattery end without anyone being hurt?
But I do think eventually that audince's suspension of disbelief will apply when they are presented with a life-like character in a compelling plot... something that is a bit more than a technology demo that the current crop of realistic digital actors are now.
not a chance, and a really long time off.
The horsepower needed to do this will not show up in your consumer level video cards ($200-$500 is consumer level... We use $3000.00 - $6000.00 video cards in the video production industry) for a really really long time.
Hell it isnt here yet for the multi-thousand dollar cards, so when it shows up here, expect about 3-4 years after that to get your hands on it.
Basically, dont wait for it dont even look for it. Just expect sometime in the next 10 years to look at a new video card and say "cool!"
Do not look at laser with remaining good eye.
Ahem, you can do real motion picture as an affordable hobby right now.
BMRT - Pixar's rendering software = $0.00
Linux Cluster Render Farm = $1000.00 (cheapie, can probably get for free as older pentiums)
Linux Non linear video editor - Broadcast2000 = $0.00
Nice dual processor Linux box $2000.00
DV video cam and hardware to get it in the linux box $1500.00
$4500.00 and Voila you have a complete system less some pre-purchased effects (Like Cult-effects for after effects) everything short of the 35mm film printer is there and ready for you.
Now the only thing is needed is talent and skill.... that you cant buy.
Do not look at laser with remaining good eye.
At first glance, I was thought this was a Tripping The Rift reference.
--
--
"Outlook not so good." That magic 8-ball knows everything! I'll ask about Exchange Server next.
I've got to wonder, though, if a CG character will ever become so popular that it will end up in other CG movies completely unrelated to the first one it appeared in.
It's already happened, although not due to the popularity of the character as such.
Pixar did a short film about an old guy playing a chess match with himself ("Gerry's Game"? I don't remember the exact title) which was shown along with "A Bug's Life". They re-used that "actor" as the toy restoration artist in "Toy Story 2".
Building even digital models is not a cheap process, they reuse them where possible. (Pixar also used the ravine from "A Bug's Life", redressed somewhat, as part of the planet of Zurg's fortress for the Buzz Lightyear opening sequence for "Toy Story 2", and many of the background objects in "Toy Story" had been in earlier Pixar shorts -- the Luxo lamp, for an obvious example.)
-- Alastair
If that were the case, then how come animation does so well? Even series/movies from american studios which tend not to have the dramatic range of anime.
I know *I* have been deeply affected by some of the anime I have watched, and I know many other people that have as well.
Secondly - your basic assumption is faulty. Even with animated works (cel or cg) there is a real person behind each character. Because whilst the image may be completely fake, the VOICES are not.
Now, given the current speech synthesis technology, if a series was voiced using it I *would* be put off. Because the voices have just as big an effect on the emotional impact of a show as the visuals. Hell, I think they have more. Many a fantastic anime series/movie has been ruined by a pathetic english dub cast.
And there is another medium for stories that has NO dependence on live people for the telling - BOOKS. By your reasoning books must be the most boring things in the world, because by definition all the characters in them are entirely fictional. And if none of the people you talked to are interested in books, they are hardly a representative sample, are they?
You think rendering times suck now? In 1992, I managed a graphics lab, where about 30 students did a bunch of short animations... Each pair of students allowed 450 frames each (15 seconds at full speed). The rendering took MONTHS of CPU time. Our render farm was a set of RS-6000s and SGIs that were near the commercial state of the art at the time (total of about 12 machines worth about $1M, retail). all of that for what ended up being less than 10 minutes of video.
The file server had a 1GB disk (5/25" full-height SCSI drive) dedicated to the graphics frames... As stuff got printed to laser disk, it was dumped onto an 8MM tape to make room for the rest.
Not that I'm complaining.. It was fun and interesting at the time we did it. We were pushing what was possible with the available tech, and I still enjoy watching the resulting videos, from time to time.
--
Sometimes boldness is in fashion. Sometimes only the brave will be bold.
Because anything fast enough to be state of the art has to be small enough to fit on one die, if you're gonna make ten VSLI chips, you might as well make ten thousand, and sell them on the retail market. Most of the cost is in making the master..
Anyone in the hardware end of things to support/refute this thesis?
--
Sometimes boldness is in fashion. Sometimes only the brave will be bold.
Look at the capabilities of "pixel shaders" before you assume hardware manufacturers _need_ to do the algorithm. Pixel shaders provide the equivelant of microcode level access to the innermost aspects of the rendering pipeline. This may already be possible today with the G3. Ask me in a week... :)
Next generation game engines will be doing this.
My problem with the characters in FF wasn't the texture of their skin. Every one of them (except the cops in mecha suits) moved like a Disney animatronic figure. This is why many cartoon characters can seem very realistic: many cel animators concentrate on the movement and articulation of the bodies, rather than detail of surfaces. When will the CG industry learn that detail is much, much less important than giving your actors believable jaw movement?
http://bbspot.com/News/2001/04/flesh.html
--
"that's not encryption - it's a new perl script that I'm working on..." - from some Matrix parody
the first time i saw the preview on TV, i thought they were real actors. i wasn't really paying much attention, and didn't think anything of it. now, this comes along... i wonder how much better it'd look in movie animations.
I think that exaggerates the GEforce 3's performance, which, realistically, is about half that, but still, that's a big ratio. Why is Wildcat only at 2.7 million triangles per second? That's way too low today.
NVidia had a "pro version" (called Quadro) and a "gamer version" (called GeForce) of each product line, with the gamer version crippled in some way. (The difference was a jumper change read by the driver.) With the GeForce 3, there's a "Quadro DCC", with dual-head output, but the performance is about the same as with the GEforce 3. It seems to be mostly a branding thing; the boards are sold through 3DS Max dealers only.
The low end has eaten the high end.
I have one of those $3000 cards from a few years ago, and the new Nvidia GEForce 3, at $250, is better. In fact, right now it's hard to spend $3000 on a graphics board and get something significantly better than the consumer products, because the high-end graphics companies are money starved (or gone) and aren't keeping up.
This photon rendering may be doable like radiosity - render the radiosity map once, then view from different angles. This allows real-time walk-throughs of static scenes.
If you had read the article, Commander, you had found out that this new technology in rendering is based on subsurface scattering.
:]
Instead of assuming that light will scatter to all places on top of the surface it also scatters inside the material.
This is, because not all materials are non-translucent. For example: milk looks as a non-translucent white, but a very tiny drop of milk is actually translucent.
The same goes for skin, Guinness and snow. By assuming that the light will not scatter back at the same place as the light impacts the material, but also makes a little travel inside the material, the overal image looks much more confincing.
Please read before commenting, Taco
This is a replacement signature.
Check out:r df /
http://graphics.stanford.edu/~henrik/papers/bss
The picture on the right has more realistic skin not because of the blemishes, but because it's more well "blended", a hint of skin oils I guess. The one on the left looks like dry plaster.
I suspect most people have an instinctive way of judging skin. e.g. healthy skin looks good to us. The flipside is that most people don't know what they are looking for or see when they judge realism - it's instinctive.
Well, this is a good thing, you know, since it allows for movies that do really cool things without the limitations of human actors. However, it does lead to a future wher we are robbed of human actors in the first place. Rummages through SF library. Lesse, I got a Poul Anderson, some Heinlein, a little Asimov... coupla Kubrics, a nice Piers Anthony or three.... Read and think, people, read and think!
If the subject is "all Greek to you", good! Take a classical language today!
Live life to the point of tears
io hymen hymnaee io
io hymen hymnaee
... because I saw the movie and all the characters definitely had facial blemishes. That's one of the things that impressed me about the graphics, not even the heroine had a flawless complexion.
---
The first, obvious application aside from games is going to be about ten years from now, when, in conjunction with your AI Design program, you can program your own porn movies on your own workstation using avatars of your fav celebrities. Insert yourself into the movie, maybe with the body of a body builder. Or even better, a novelty program sold to the consumer where they can insert the avatar and you can interact with the players in the movie, directing the action. You know this is going to be a great busines opportunity for someone, someday. The MPAA and the RIAA are going to freak. (all that unlicense use of celebrity images)
If anyone is using a workstation by then.
"It is a greater offense to steal men's labor, than their clothes"
It seems that they've replaced the files with small useless ones to avoid being slaughtered by the /. effect. (I'd imagine it would be powerfull with all of us downloading 8 meg movies)
Alcohol, Tobacco and Firearms should be the name of a store, not a government agency.
There was a movie called Looker in the mid 1980s. You might still even be able to find it in some rental places.
In the movie, the computer graphics technology develops to the point to where actors are unnecessary. They can render perfect models of actors. Therefore, in order to have a monopoly, they start killing off all the "perfect" actors and actresses.
One laughable aspect was that they rendered the actors over a real background. A camera would pan across an empty set, then in post production they would insert computer generated actors. The obvious flaw is that you could render the background set much easier than the actors themselves.
--
"Linux is a cancer" -- Steve Ballmer, CEO Microsoft.
I'll see your senator, and I'll raise you two judges.
In an interview about Shrek, they said that they sorta did subsurface scattering. What they really did was build several polygonal skin layers on top of each other rather than simulate the actual sub surface scattering. I can't get to the site (/. effect probably) to read the actual paper on the new subsurface scattering method to compare it.
In the current Cinefex issue, they talk about the FF movie. It is said that they actually had the characters looking more realistic, but felt that they needed to tone the realism down because it freaked people out to watch even more realistic characters have less than perfect animation. Actually, the same interview about Shrek mentioned about also said that they toned down the Fiona realism because it didn't stylistically match the rest of the film.
I'm a loser baby, so why don't you kill me.
Funny, yes. Informative, no. It's a porn joke you moderating genius you.
What? Zits? Pimples? Moles? Scars? Speak for yourself Commander! My face is smoother than Hillary Rosen's ass, just without the smell. ;)
Oh, so that's how they did it! Gosh, movie making technology is getting more advanced all the time!
main(c,r){for(r=32;r;) printf(++c>31?c=!r--,"\n":c<r?" ":~c&r?" `":" #");}
As I recall, in that article, the potential pornographic uses of improved flesh rendering were emphasized. Let's face it, the porn industry will be some of the first people to deploy this technology large-scale. They pioneered BetaMax, VHS, GIF, JPEG, internet multimedia as a whole, and this too will be lead by the noble pornographers. God bless them, every one!
USA Intellectual Property Laws: 5 monkeys, 1 hour.
I'm the stranger...posting to