Disney's Hyperion; combining the best bits of caching with path-tracing

The approach is really interesting, instead of just firing rays at geometry, they employ the idea of firing bundles of rays to, in effect, get the benefits of caching techniques without doing any actual cache/pre-processing phases.

The bottom of the article says everything in a nutshell, the engine they built is actually superior at scaling up compared to the Arnold engine and even kicks it to the moon in terms of rendering speed in complex textured scenes. It actually looks like a person with consumer-grade hardware can now achieve photoreal animation with ease.

They also have developed a new SSS algorithm which hasn’t been revealed yet, I wonder if this new approach to lighting can be applied to Cycles because it doesn’t seem like a major violation of its paradigm if there’s no actual pre-processing involved.

I wish the amazing tech was being used for something that wasn’t quite so… Disney.

The last few Disney 3d movies have been excellent; animation, storytelling, look & feel – all have been really, really good. Seems like Disney animation has been taking off, while Pixar has been just, meh. No doubt Disney has benefited the most from the Pixar acquisition.

Same old unoptimized/exaggerated comparisons you see from everyone in the rendering game. Ray bundling isn’t a novel approach, really. People hoping for photorealistic animations on home computer hardware are living in a dream world.

When I read that article the sense I got was that the kind of scenes this engine excels at rendering are exactly the kind lone wolfs or small teams are unlikely to produce. Lots of geometry lots of textures.

Off-topic: False :wink: Wreck-It-Ralph was great, but Frozen was GARBAGE. The animation had lots of appeal, the look was great, but the story was BLAND. My kids keep wanting to watch it, and my response is always, “…why??” Monsters University, on the other hand…

Butthurt Arnold fanboys to the rescue! There’s a lot of novelty in this method, which you’d know had you cared to read the article.

Of course the comparisons in the paper are bullshit. But of course, it would not be real computer science unless you can show your method to be linear (or even constant) while competing methods are shown to be exponential.

And before you go and call me a Pixar/Disney fanboy, I’m just gonna go ahead and say that all Pixar movies are terrible (and so is Renderman).

You know, it’s really hard to judge how photoreal a rendering engine is when the example is a cartoony plastic-looking disney type animation. I’d like to see something like a human face that has a real-world analog to compare to.

Frozen was not GARBAGE! Especially the story. MU’s story, on the other hand, sucks. It’s like:
“Hey, what should our next movie be about?”
“Ehhh… A, remember that blue-tall guy and his small-green friend? How about a prequel?”
“A prequel? About what?”
“I don’t know… School or something?”
“Seems good…”

BTW, nice renderer, looking forward to BH6. :smiley:

Wasn’t one of Arnold selling point being able to deal with mammoth assets and textures easy peasy, increasing A LOT the complexity not hurting performance that much…? Those benchmarks of 1000+ minutes vs 11, err…

“Once a ray has done that, that last ray remembers where it came from so you start populating backwards where your first bounce light sources will be and so on.” This means that the next ray that happens to land in the same place going in roughly the same direction will be informed where the illumination sources will be.

Isn’t this a similar logic to http://cgg.mff.cuni.cz/~jaroslav/papers/2014-onlineis/index.htm i.e. informing the ray with previously taken data?

I will throw my stuff in there too :smiley:

First the optimizations made here are neither for a gpu or any sort of a home computing level but on a server farm level.
Secondly the renderer is meant for scaling with texture data, and NOT with ray shooting itself, which does NOT increase baseline ray speed (ray grouping is only for texture/geo streaming efficiency!).
Third, the out of core memory optimization is SSD dependent and will speed up only very huge texture files which are not intended anyway to be in main ram.

So yea, this is NOT a home artist ultra-wonder-i-1337-all-hollywood-in-the-dust-engine, but an efficient renderer ONLY for RENDER tasks at HUGE geo/texturing scales and NOT interactive like cycles, lux, oct, arnold and other renderers in the viewport even, but similar to Weta Manuka or PRman Server Pro.

Meaning that you need to pass the scene like in good 'ol days to the render scheduler and you see how it renders (fast) but then again at very high complexity of billions and terabytes of poly and textures. So yes, no viewport interactivity. And i believe this is where (for preview) PRman VCM comes in, like Weta’s new development Gazebo for Viewport and Manuka for huge scale render only.

Maybe the next engine for avatar 2 ? :smiley:

Also i need to add that although there are a lot of SSS algorithms, most of them are still a cheat, due to the missing underlying bones,tissue layers, muscles, etc… each with their own densities in many productions. Maybe Weta with their tissue system have that (but even that i doubt). SSS works well in homogenous media like wax, milk, etc. but we while we simulate MLT nowadays, we are still far away as an industry from realistic creatures with full underlying tissue and bones(weta’s tissue is an exception and not the industry as a whole). So pure technically speaking SSS is still far away from unbiased calculus cause it also needs more underlying geometry with densities for characters!

If i am wrong please feel free to correct me.

I wasn’t thinking about the broader quality of the films, I was thinking specifically about the art. For example, the image of the female used in the article looks nigh indistinguishable from Anna, Elsa, Rapunzel, and all female characters contained therein. This of course brings up the comment by Frozen animation director Lino Disalvo, which he will never live down:

"Historically speaking, animating female characters are really, really difficult, ’cause they have to go through these range of emotions, but they’re very, very — you have to keep them pretty and they’re very sensitive to — you can get them off a model very quickly. So, having a film with two hero female characters was really tough, and having them both in the scene and look very different if they’re echoing the same expression; that Elsa looking angry looks different from Anna (Kristen Bell) being angry.”

Gotta’ keep those women pretty and saleable. In the case of CGI, that means they all look freaking identical.

Similarly, noted CGI curmudgeon John Kricfalusi has often said that he doesn’t like CGI because it all looks like it’s made from the same stuff. That’s not surprising since they are all using the same math. I loved Paperman’s look because it looked like it was made from different stuff. Disney won’t even give us different styles, much less different stuff.

Cloudy With a Chance of Meatballs, Hotel Transylvania, Flushed Away, even garbage like Igor. It’s the same stuff, but it’s at least a different style.

Regardless of the quality, I’ve grown so utterly tired of the Disney polish. It looks nice, it goes down easy, but after thirty years of eating this shit, it all tastes the same.

Well it does appear to be true that some key materials like human skin have looked more or less the same since Toy Story, the only difference now it seems is the more physical way they do the speculars (and maybe some improvements to the SSS).

Frankly, the only animation outfit that seems to try major differentiations in the overall look of the CG is Dreamworks, sure there have been clear advances in the realism of some materials as well as the complexity of the scene, but it seems to me like the present-day 3D cartoon look is here to stay much like the long life of cell-shaded 2D animation (which stayed more or less the same for over 50 years until the advent of computer animation software).


The thing is that the “same stuff” style comes from the fact that nearly every studio tries to keep the main characters as clean as possible,with clear eyes, and elastic body material (stuff) for extreme emotions, the reason being that children (main target of big productions) react much, much better to clear expressions uncluttered by too much textures/displacement. This of course with smooth subd’s forces that “stuff” feeling across all these years.
Not to mention that you win of course less render time, less modeling, clearer poses/expressions which is convenient.

An exception would be for ex. Avatar. But then again it was not a PG-0 Movie and it was “molded” on reality not a pure “invented” CG Animation.

The more realistic you get, the more different “stuff” you get, based on the material and texture. BUT the more realistic you get, you need to be very careful to not geo/texture overcrowd, especially the main characters, else you get into the rightfully feared uncanny valley, and that is what the main studios want to avoid at all cost, even with the risc of being stuck in the same “stuff” (which they are :-P).

Like Ace said, this will stay a looong time, but parallel there will be (is?) more direction towards more detailed textures and geo which will make other “stuff”, not only style. Still these movies won’t be for 6yo, but for older audiences and are right now only in their infancy.

Here i speak about pure CG, NOT creatures like the Rocket Raccoon in the Guardians, which is actually in the VFX domain.

Besides that, I will probably see Big Hero 6 in theaters anyway because it looks like a fun movie (much like Wreck-It-Ralph and Bolt before it).

I think Disney is nailing the formula for what a good popcorn movie should look like.

I disagree that the cel-shading of cartoons all looked the same. Especially the backgrounds, which could be made with sponges, brushes, oils, watercolors. There was far greater variety in the old hand-drawn styles. Again, I don’t want to seem too much of a Kricfalusi acolyte (his blog is amazing, though), but he spends a lot of time on the different styles and textures. Ralph Bakshi is in another world than Disney. As is Hanna-Barbera, or Fleischer. Even Disney changed it up. Sleeping Beauty looked very different from Pinocchio.

A good read that sums up Kricfalusi’s thoughts can be found here: http://johnkstuff.blogspot.com/2008/08/artistic-graphic-shading-in-cartoons-vs.html

If anything, the massive power provided by computer animation should make them greatly expand what they animate, not simply allow them to do more of what they were previously doing. They could easily make a film that looks like a living oil painting. Look at the truly awe-inspiring art direction in What Dreams May Come. That has nothing to do with realism, it has to do with being inventive with the artistic tool that they have.

They have no excuse other than pure commercialism for their derivative and tired style. That’s actually fine. They want to make money, people eat that stuff up. I understand their motivation. Art and money rarely line up, and as none other than Michael Eisner said:

“We have no obligation to make history. We have no obligation to make art. We have no obligation to make a statement. To make money is our only objective.”

Just as he has no obligation to make art, I have no obligation to respect him or his legacy.

I actually complain about the same stuff in video games. Just as we have an arms race of sorts in movie CGI technology, the big budget games all feature the same crap. And yet, those are the games that sell a bazillion copies. Interactive art like Okami, Shadow of the Colossus, and Jet Set Radio sell poorly, while Grand Theft Duty XII makes more money than the GDP of Sub-Saharan Africa. I do appreciate the pull of commercial success. But for all except the most hard-core Randians, earning money has never been something that is itself worthy of praise.

Until a Disney princess movie for instance doesn’t followup with millions of little girls squealing for more of it, there’s no real reason for them to change things up (that goes for the plot styles as well).

Take Frozen for instance, the lines to see Anna and Elsa at the Magic Kingdom was 2 hours long and every Elsa Halloween costume in sight has been sold out. Elsa herself is actually one of the most popular Disney princess figures of all time. Meanwhile you have cases where someone actually tries to make a moving work of art and it doesn’t turn out so popular once you find that the average person doesn’t get it. People in general want movies where they get the plot, they get what’s going on, and they can follow it through without popping a bunch of neurons.

It’s your choice if you would rather want to watch a movie that looks like Leonardo De Vinci and Vincent Van Gogh in motion rather than something that feels like a 90 minute amusement park ride, same with those group of people who argue that the brush and canvas is the only true way to do art and all your CG work is, at best, a cheaply made imitation.

I couldn’t say it better. And although i really think the last comments about “stuff” have nothing to do with the topic about the renderer, i want to say that there are some clear rules with 3D that do not necessarily apply to 2D. You could do anything in it of course, BUT if you want to make it accessible to most people or excite them, you need to stick to the mainstream even in art. There are things that work and others that don’t work. The same goes for Stereoscopic. The more dimensions you have, the tighter your rules are.

The more eccentric the (3D) art, the less viewers you will naturally have, and there are one zillion reasons (and art books) why. Putting the discussion about the story itself apart (there is a difference between art and technique!) i think even with the actual “stuff” renderers you do real art. Not to forget that art is also like taste, subjective. Technique is not. So there is naturally no room to produce something (on a high budget) that MOST of the people won’t like. And even that is often a guessing game no matter how much market analysts you employ.

Art IS entertainment, it was from the oldest cave paintings and fire shadows to the newest CGI.
If entertainment also employs philosophy in its story is another thing by itself.
And regarding money: if its not art to most people, they will not buy it. It doesn’t make it “not art” but also neither means the “for money art” is not art. :yes:

To add, movies that try to be a work of art can even lose the CG-centric crowd if done the wrong way or in a too confusing way.

The BF actually saw this themselves when they released Elephants Dream, in a way it could be seen as more of a living work of art rather than a standard film, but the plot was more or less completely lost on the dozens of people here on BA at the time (then Elysiun) who were busy trying to figure out just what it was trying to say.

That said, I’m not opposed to movies that would be like living art and I’m sure that one done right can find success somewhere, just that it’s to generalist to say that all movies done in the standard animation these days is a sign of greed over creativity (that quote by Eisner for instance, one needs to note that Disney actually had a number of box office disasters in terms of animation during his later years, so the average person is not so much like livestock that he will see and buy anything just because of the Disney logo).

Since I’m really familiar with 3D CG, it’s difficult for me to watch one of these movies and separate myself from how it’s actually done. Most fans of the Disney/Pixar/Dreamworks/ movies don’t know and don’t care how it’s made, just that it appeals to them in some way. Wreck-it Ralph was a great story, great animation… just a fun movie. I didn’t care that the SSS was off, or that the global illumination could have been better (not saying those two things were off); the point is that I was entertained.

If you watch Tangled closely, you’ll notice that the animation is very very close to the style used in classic Disney cel-based films. This is intentional, and that’s why Disney is the best at animation and storytelling. They are the best and have been for decades. If you don’t believe me, just check their box office receipts. You may hate Frozen and are sick and tired of that dumb song, but it filled Disney’s pockets.

The point is, you have to look beyond the “plastic” skin and watch how Disney animates those manikins and breathes life into them. That’s where the magic is.