Framestore Creatures & Houdini | Framestore | Character FX & Crowds Production Talks

[Music] next up we've got a med graph he returned to frame store as joint head of cg in 2017 after a brief absent absence working on Star Wars the last Jedi and ready player one is lead effects artist at ILM Ahmed's frame store credit lists include major advertising hits such as McDonald's reindeer ready Sainsbury's mugs Christmas calamity and shell v-power shape-shifter over to a med to take us through some work at frame store good evening everyone and thank you for coming thank you for sticking around thank you for Chris firm and side effects for inviting me along to speak to you all today as Chris said my name's Ahmed I'm one of the heads of cg at frame saw here in London and today I'm gonna talk to you through some of the creature work that we've done in the department and I'll explain how we've used Houdini to solve some of the challenges that we've solved along the way I'll first start by telling you a little bit about our department right so historically we were mainly a commercials department of producing some really iconic and memorable adverts but over the years our department has evolved and grown exponentially both in in size and in our offerings so we've grown and adapted with the new and emerging technologies and the way people engage with digital content so we've got three main offerings televisions kind of self-explanatory immersive we need a whole other presentation to talk through that and creating theme park rides from start to finish VR and AR experiences and of course we're still creating commercials I'd like to add that this year we've been fortunate enough to have ves nominations in each of our departments offering so that's a really really good achievement that we're all very proud of all right so that's enough about the department let's move on to our main topic a few years back we had the opportunity to recreate this classic British character her name Smaug was kind of a stylized cartoony cat if no one's familiar with it the spot was three and a half minutes long and it featured this fat cat when the job was first awarded we had big discussions among the team about how we were gonna rig and deform this cat and the the more we talked to her the the more I was convinced that we should be simulating as as much of the surface of the cat as we can it's this big blubbery cat and you know there's only so much you can do with blend shapes a few of the guys argued that we should spend more time adding more sophisticated rig controls belly controls so the animators can move it around and all of that stuff I wasn't convinced in my mind's eye you know a real cat can't animate its belly this way or that way right that's that's all driven by physics so that's that's the road I wanted to go down and around the same time luckily for us Houdini had just shipped with this brilliant new solver the FBM solver and I was really keen to try it out as it was it seemed like a very good way to simulate fleshy and fatty materials so for me seeing the simulation was really exciting for many reasons I did this really early on in the project just kind of as a as a proof-of-concept just to make sure that I had in my head which would actually work so I wanted to create an FIM volume which had an inner skeleton and have that skeleton drive everything else around it so that's that's essentially what we've got here right with we've got this animator tube that's just kind of bobbing up and down which can represent our skeleton when we have the sphere that's wrapped around it and all the points that the tube is intersecting with the sphere that's just constrained on you hit simulate and you let go and it does some really fun stuff and because it's in Houdini you know you can manipulate everything to a finite level so you know you can change the properties on a Parrothead basis giving more control over the behavior and introducing some randomizations etc to give it a more organic look here's a bit more testing some early R&D I don't know if you can see on the screen there's some some green points along the the front and the back so what were you can't see is then an invisible skeleton that's kind of driving that simulation so the same as the clip that shows you earlier there's a skeleton that's animated that's inserting intersecting with the geometry and that's driving the simulation everything else is is just driven by the FIM physics and I've added a bouncing brick in there just for fun here you can see a side-by-side comparison of what's coming straight from rigging on the far side and once it's gone through the FIM solver so you can see it gives us a lot more kind of organic deformation as well as volume preserving because of this on this particular project our rigging team just didn't really have to think about skinning the character much at all you know spending time painting complicated skin weights and crap clusters and all of that kind of stuff so it freed them up to spend more time on the actual mechanics of the rig to create something that's really good and robust for the for the animators to work with so during the project we created this fun little animation test which kind of evolved and changed over over the breadth of the project initially we use it internally just for testing or aspects of of the assay and modeling the rigging testing the animation shaders looked at all of that stuff but we also use it to present a client there's something there was a bit more fun than you just standards turntable just a you know classic static model spinning around here we can see that same animation just with Fe M sim running on top of it's a bit creepy as you can see we we ended up simulating much much more of the cat than say compared to the the test with the bouncing brick that you guys saw earlier the more we ran shots and the more we did tests I just kind of slowly slowly side encroaching on more of the character and until we're pretty much simulating everything it said the the tail the paws and the face and ears we were even simulating the cheeks up here which gave us some really nice secondary animation here's another fun clip I must admit I turned the settings up just for a bit of a giggle but I mean the the intention was really let me backtrack as I mentioned earlier the the spot was three and a half minutes long and we had the lot of shots for that that had this cat in it and we had a very very small CFX team on the on the project ie just me so [Music] I wanted to create this automated system I didn't want to go into every shot and run simulations and update caches and all of that stuff so I wanted to automate as much as possible I had this one master scene which I updated throughout the production and that was triggered on the farm at the end of each animation cache coming out of Maya this then simulated these CFX components including some first simulation and at the end of that spat out a couple of QC renders one with some really cheap fur and one of just the skin so we can see the how the skin is moving in the forming and that cut out you know a whole process of an animator creating a cache and going up to a lighter and saying hey I've got a cache can you push it through and render so this you know it was it was really good and and liberating for everyone on the team and I also gave some control over to the to the animators again like I say I was just a one-man team and I didn't want to animate things I didn't didn't really have time to do it so I created this squishiness attribute which was exposed in the my rig which the animators can animate within their shot so you know if the cat is moving fast and running around then they can tighten it up and all the skin would would come together if it's sitting down on the table or if it's relaxed they can increase the squishiness and just kind of let its flab settle a little bit here's one of those QC renders I was telling you about this is actually one of my favorite shots I love the animation on her face and that same animation is accentuated by you could see their faces getting squished against her arms had her cheeks against question like that and that's coming purely from the simulation which which was great we didn't have to do blend shapes and all of that complicated SEF and here's a snippet from the final project [Music] [Music] [Applause] [Music] all right next step is a project that we did a couple of years back for Mercedes we were tasked with creating this photo real blinds head which we then tracked onto this high-powered alpha male businessman previously we'd used in-house fur tools for all our grooming needs which was developed by the film department but around this time that we were ordered this project Houdini 16 had just released with this whole new set of fur tools which we were very keen to try out very quickly who we found that the new fire system was very powerful for the simple fact that it was in Houdini meaning it was old node-based it was all procedural data can be manipulated anywhere and transferred as you want in the scene helping us create exactly the results that we needed so for instance we it was possible to create some really intricate sub clumps and fine-tune how the fur looked to create realistic shapes and variations one of the biggest challenges that we faced was dealing with complex first simulations and collisions we had a couple of shots where the the guy was passing his hands through his hair and you know that's that that kind of that kind of stuff is quite tricky of course we were using the wire solver which was wrapped up into the into the fur tools at the time and you know for those of you have used the wire solver you know that can be quite fickle eventually with enough substeps tickling the right parameters and a few secret handshakes we we managed to get a sim that worked here's a intermediate render of the same thing that you saw we can see some hands intersecting with the fur and here's the final thing [Music] [Music] [Music] mood changing technology at the touch of one button [Music] next step I'm gonna talk about this project that we did for Sony a7 camera so earlier last year we delivered this project and we were tasked with creating this heard of black wildebeest running a super slow motion initially when we got the script and we awarded the project the our main task was to embellish some live action plate so we were gonna go out on a shoot and actually capture lots of footage of these wild wildebeest running around in super slow motion with phantom cameras and all of that stuff and we were gonna add in some extra CG wildebeest just to bolster the herb maybe some close-ups of like hooves landing and that kind of stuff so we went out on the shoot to shoot the worldA beast not with a gun but with a camera and they they weren't there we turned up and the guy that owns the ranch she was like oh yeah all the all the world to be so present the miles and miles away we can't get anywhere near them sorry here's an empty field so suddenly while we're on shoot the brief completely changed and you know the client still wanted to have this this idea the black wildebeest running at super slow motion so we have to think of really kind of quick and efficient ways to still deliver the brief I mean we still had to create the wildebeest anyway but we needed to build it in a much more kind of full frame type of way here here of a couple of breakdowns of the shots so the premise of the ad is this famous footballer guy is out on safari taking pictures with his brand new a7 camera and the pictures that he's taken are so detailed and so sharp that he's transported into this dreamlike state where he can walk through the head and into the picture in super slow motion each shot essentially ended up being kind of close to a still image where our CG renders were moving at around 4,000 frames a second here's an animator created this extended run cycle a range of motion if you will which we then use and stretched out 4000 frames a second and within Houdini we created this kind of fake crowd system it's you know just a simple point instancing where you can scatter points around and you choose exactly the the point along the the run cycle where you want it to have on the screen of course these characters had really long flowing tails so we had to create some first simulations because the director on this commercial was constantly making changes to the layout of the head up to the up to the last day and adjusting which parts of the run cycle which were being used by which wildebeest on which part of the screen we never really knew which parts of the run cycle we were gonna see in the final image which meant we had to create a first simulation for the entirety of of the run cycle but there was no way I was gonna simulate at 4,000 frames a second I would have meant I would have been simulating at over 24,000 frames which is just crazy so after some experimentations I settled on swimming at around 250 frames a second which you know it had enough detail and subsets that allowed me to time stretch it and then I added some extra noise and deformations and post presses and that kind of stuff to just add a bit more interesting look to the final image I also like to render my test renders with bright neon colors just for fun speaking of bright neon when we delivered the job we were trying to think of a cool way to present our work that we've done this this shot is is it's nothing to do with the commercial this was just kind of a bit of fun that we that we did internally so we have this fun run cycle which we created with the intention of only being viewed as super slow-motion so you know I'm sure you can spot some clicks and pops and that kind of stuff but we didn't carry this this was just for for us internally for fun around the same time that the project delivered the Eurovision Song Contest that had just ads I don't know if any of you familiar with that but anyway one of the acts apparently had some really impressive stage lighting which inspired grant Walker co-head of CG to recreate them in CG using a bit of chops in Houdini we put together this slideshow for our wildebeests to run through all right last up is this commercial that we recently delivered just before Christmas last year for McDonald's we were tasked with creating six different reindeer to fly Santa around for the night while he delivered presents and ate lots of mince pies we had about three and a half to four months to deliver close to 40 shots of full CG reindeer and apart from creating the reindeer we were you know we also had to create all kinds of volumetric effects and you know they've got this rigging around them and all kinds of other stuff but I won't touch on that too much today right so we kicked off with one of the most important things if not the most important things when creating a creature in CG or or creating anything in CG for that matter collecting and studying reference so I firmly believe that in order to recreate something from from scratch for us in CG we should try to to understand and study as much of what we're trying to create as possible so we steeped ourselves in reference we visited a farm where we spent some time hanging out with some reindeer feeding them stroking them taking lots of pictures we had a TV mounted up in our office that was just constantly looping through clips and pictures of reindeer we were just like reindeer crazy in pre-production we started by cranes some sketches and concept designs which slowly evolved and changed and you know helped us answer some of the the broader questions very early on like size and shape of the reindeer their color the mechanics of how they're all going to attach to the sleigh and how many reindeer were gonna have this was this was a hot debate at one point within the office should Santa have six reindeer should something have eight reindeers it was a bit divisive but we we ended up with six thankfully would have been too heavy on our render farm otherwise it was important to figure out a lot of this stuff before we got into creating the assets and pre-production for me is always quite fun stage so right from the start we knew that we wanted to leverage Houdini as much as we can with the data coming out of Maya so we decided to use FBX exports it was easy enough to get our animated joints out of mind into Houdini and then we had full skeleton rig and we used that as our basis for driving a lot of the character stuff so using that that same FBX skeleton they just saw we built a Houdini muscle rig around it was a first foray into using Houdini muscles and there was a lot of learning to be had we ended up using just the the simple jiggle controls to get momentum and movement into the muscles rather than using full dynamic simulations at that stage due to certain bugs that we discovered and time constraints as well what what this gave us was was good enough for what we needed we then wrapped those muscles up in a Fe M solid embed object to get more physically accurate simulation so this this approaches it's not too dissimilar to what we saw on mug the the fat cat earlier it's it's kind of an evolution of that process it's just slightly more sophisticated so we've got before where we just had a skeleton driving everything now we have the skeleton which was driving some muscles and those muscles had attributes and parameters which were then driving the the FE M on top of that and then on top of that simulation we sometimes added another layer of detail to certain shots and some some of the foreground characters so using the the new hybrid object which has recently become available to us we were added to add some fine wrinkles and detailing in the skin like skin sliding and whatnot so onto grooming the fur groom was one of the biggest challenges for us from both an artistic and a technical point of view the approach for creating the base guides relied quite heavily on manual combing and masking so combing and sculpting each of the guides adding curvature volumes and flow just carefully to each section of the reindeer we we had very specific reference that we were honing in on and we were looking for the exact shapes and flows as you know manual combing is was essentially one of the only ways for us to achieve that I know it's a slightly destructive workflow to be sure but at this stage allowed for a level of detail control which would be very difficult to achieve procedurally right so after scouting the base guides we moved on to then we moved on to a more procedural workflow and relying on several levels of sub clumps we used sculpted guys to generate the first layer of sub hairs just to kind of initialize the system this also dictates the overall clump size a second layer of sub hairs which is denser than the first one is then clumped on top of that and and so on you know it's it's an iterative process so you can just kind of keep honing in on as many sub clumps as you need different noises are obviously applied along the way to kind of break up the clumps and add some more details and naturalistic looks here's more of the same thing you can see it's yeah that's the sculpted guys some sub clumps and some extra sub clumps here are a few whip groom renders so adding small details like clumps along the mouth to simulate the the wet fur and breaking the symmetry along the face and the foreheads all helped to add to the to the realism we had some really close-up shots of the reindeer where the camera was gonna get very very close to the eye and and and the muzzle so we had to make sure that our assets would hold up to the camera again so close so you know we we did the same thing we stuck cameras all over the the asset and we'd hone in and we'd scrutinize and add in as much details as we can you can see us so yeah when whenever we're presenting renders were we're always checking the reference in there so you've always got a frame of reference like this is the real one this is our CG one what what does everyone have that ours doesn't and you know constantly it's raising constantly just comparing and and and trying to hone in on on what nature just creates by itself right so in real life brains eeeh grow very antlers around kind of late spring they actually grow the the antlers that at an astonishing rate they're when after the the antlers fall off and they're growing them again they they grow at about an an inch a day which is crazy and then around August time that fairy layer of skin begins to rub and peel off and it can look pretty gruesome with like blood dripping down the antlers and down onto its face it's loose them yeah we saw some crazy pictures either way by December time and Christmas generally the real reindeer the the antlers would completely lose their fuzz or they would have fallen off altogether but you know for this project we decided to take some creative and artistic license you know if everyone likes a bit of very very antlers so we decided that our reindeer were gonna have furry antlers and you know who's to say that Santa's reindeer on a different breed of reindeer they keep very antlers throughout the year that's this so we had six different antlers to make fuzzy but eight for each one of them because the the shapes varied differently the we didn't have the same topology so they did there was a bit of a manual process just to get the the grooms on on all of them but since the fur was quite short and messy just using curved flowers to draw draw directly on on the geometry proved kind of a very effective and quick way to draw the different measures so adding extra parting lines helped to add more details along with some procedural noise after establishing the workflow and deciding on on the look of one of the antlers it was pretty straightforward to recreate on on all the other ones it's like fitting minute job so this this for me was one of the one of the most interesting parts of the job so while studying grain deer we noticed that the different coloured hairs generally had different properties in length and density and thickness and and so on and you know depending if it summertime or winter time they generally had a different groom so we decided that when it came to texturing the reindeer we would drawn on how it is in the natural world so we we we decided this new approach where that we'd not tried before it was a bit experimental but we were confident that we were onto something that would work and would add a really good extra level of realism so instead of painting color maps in kind of a traditional manner and say Mari or Photoshop or whatever we decided to drive everything using painted attributes directly in Houdini the color attributes initially were linked to the hair generated so we had hair gems for each of the different color types and each of the different fare types on the reindeer it turned out that the painting those color mouths were it was actually a quite a fast way compared to the traditional way of painting textures because that's one thing we were worried about we're like is this are we going down the right routes is this gonna add extra time but actually in retrospect it was much much quicker especially when we had to create six different color patterns for each each of our six reindeer so yeah you can see here there's different different combinations and different masks allowed us to mix between the different types and the different colors of hair pretty quickly here's an early turntable render of one of our reindeer we use Arnold's we predominantly use Arnold in the department we do use mantra for some of the projects but we rely heavily on Arnold and so the same attributes that I was talking about earlier that we painted for for the different types of fur we use those exact same attributes to drive the the melanin attribute and the in the Arnold fur shader so again we're not driving anything by color it's it's all driven by attributes which you know in in the real world your your hair color comes from melanin unless you dyeing your hair purple or green or anything else you know it's it's it's a natural occurring thing so the the more we look to nature to to help us drive what we what we're doing I think the the more realistic our CG renders will be that's a close-up of the reindeer again I mean you know going back to looking at some of the the previous creatures that I showed earlier say Marg or the world of east or or the line whenever we zoom in on it on a close-up of the face and stuff the the fur usually has this kind of painterly look and you can see where the colors are blending in between between each other but we just we didn't struggle with that at all in this project for the simple fact that we just relied on driving everything through attributes and and Melanie next up first hims so first Sims for us was it was quite a sticking point ideally for me I want an effects artist or or a see effects artist to simulate the fire guides approve those and then pass those over to whoever's gonna be lighting the creature they get the the deformed guards that get the static groom that add them together everything would deform and you'd have some really cool renders you know and and there there is an option for that to do that in Houdini and the guide deform you know you can capture the the guides and and use that further down the line in the guide to form but certainly for the number of hairs that we were working with capturing the guides at third-generation time it just it wasn't working it out endless hours and hours and hours and memory limits and all sorts of stuff we just couldn't do it and even when we trimmed the fur down so thinking alright let's let's do it in chunks once we got into the guide deform stage the deformations were just were not as clean as we wanted you know guides were affecting some fur that he shouldn't have been so we eventually gave up on there and we're just simulating the fur directly in you know in in the hedge ends and generating full fur caches of just the parts of the reindeer that we wanted simulated and it looks really weird they kind of looked like Amish reindeer I always thought so that's that's what we did we just you know eventually we ended up going right we'll just simulate the longest hairs the bits that interacted with with the harness you know the harness was simulated as well we we use bullets over to do that [Music] but yeah first simulations so off the back of that I mean you know we we did what we can after some grafting we we got the fire sims out but looking to the future we've recently hired in a really clever guy to help us with some R&D work he's sitting in the crowd and he's been asking some really poignant questions he's been back and forth with Kai and he's essentially created a new guide deform for us frame saw which which were using and it's an it's super fast and efficient and it does exactly what we want and he's now he's he's also created a new new way of grooming the fur which again is faster and next up for us where we're starting to look at some feathery stuff so you know maybe next presentation we can talk about that anyway so here's the final result and now sky one presents from McDonald's [Music] [Music] sorry guys no carrots again [Music] [Applause] look no carrots [Music] they didn't leave any oak [Music] [Applause] [Music] I'm tired too you know [Applause] [Music] happy Christmas [Music] [Applause] and I'm afraid that's it thank you very much [Applause] any questions any questions beautiful work by the way thank you and kudos to the team that worked on a lot of them are in the audience of thank you guys I was wondering when you said you were painting Maps for the reindeer for the color how did you deal with the route in the tip and then also I noticed some kind of random variants between there like Browns close-up Browns and whites how how did you deal with that Ruta tip was was dealt with at the shader level when we were mixing between the the colors so we essentially we said all right so let's say for example white fur is gonna be 0.25 brown fur is gonna be 0.5 gray fur is gonna be 1 and then we we paint in that we mix and blend the colors and we normalize that so I mean anything that's that's in between just just picks one of the colors basically so you know you never really get a blend between the two it's it's either gonna be brown or white or white or brown hello how long did it take to to do the grooming of the reindeer like how long did it take the whole thing to be done but the simulation part just the chrome in the drawing the guide and roughly about 10 weeks yeah thank you hi Simon other questions all right thank you very much good thank you Thanks

Channel: Houdini

Views: 20,978

Rating: 4.9566002 out of 5

Keywords:

Id: _JQ673B6Y70

Length: 42min 37sec (2557 seconds)

Published: Mon Feb 04 2019