 Hello, everyone. Hi Hello, my name is Corrado Piscitelli. I am the technical director of my entertainment a studio animation studio based in Naples, Italy This is the making of our third animated feature film. Yeah, and Lenny the walking Liberty and this year in particular marks the 10 years anniversary of our using blender for making our movies and our cartoons and our Feature films as you can see here. We have Many projects that we and even more there are some projects that were done in the last two years But mainly I would like to point your attention to the art of happiness Cinderella the cat and yeah, and Lenny the walking Liberty. There are our three feature films They were mainly done with the the same team and so it is a natural flow of our image of our narration and Whatever we're trying to Narrate in this world What this panel would be about well, it would be about some general pipeline stuff because This was the first movie that I have Been in charge of the technical direction I started with the art of happiness where I was mainly in charge of everything 3d about that movie That since it's a 2d movie. You might understand that is very little Then I worked on gattas and rent or it's Cinderella the cat that was mainly more massively 3d in its use and More massively blender in its use and I was and I and I made most of the grant work of the 3d stuff and so I had a particular experience in this because This is even my 10 years in using blender with our feeder films. So what happened it happened that I have Particularly Many know how about how we do we did all our movies So I'm trying to talk to you about some pipeline stuff of how we do it What is our method and what can be done better? there is something about some rig pipeline stuff because we use heavily rigify for our feeder films and We found more ways to streamline and use it We're talking about working with legacy Blender plus and cons because This movie was made in Blender 2.79 B not 2.80 and I'm going on about it in a second and How we upgraded our feature workflow. I would like to show you a trailer Most of the Thank you. I hope you liked it This is our team. It is all the the group that work on that movie and I would like to show you a little bit of a Montage about our working unit. Thank you again This panel is going to be mainly about my personal experience in doing this movie and since as I said before this is the first feeder film in which I was called in a Technical direction position. So I had to plan everything All things That I didn't have to personally do on our previous feature film Cinderella the cat But I wanted to bring all that method all that know-how into this one and try to Have the best outcome possible so Our feature films as you can see from this handy graph that's over there Take generally around two and a half years to make and you can very broadly Divide them into three steps. The first step is the creation Step where you make all the concepts and you create all the assets And you need to use in the second step that is a truly layout phase where you put all the stuff Inside the same File and then you can animate it and it all comes down on a third step where you render and Composite everything we use for our feeder films heavy Compositing to to get our our look of distinctive look Let's look at the first step the concept and asset creation phase Here we had to Build basically a whole jungle. You might understand that What that means? It means we had to make lots lots lots of plants and Many times over because we started with the previous method. We use it on Cinderella the cat where we modeled everything using only vertex Vertex paint because that was light. It was it was really lightweight. It was really easy to to model things that had the A fresh look but that proved to be a little A little less Graphic for what we were trying to achieve because we needed something more worn out more Destroyed because this movie is set in a post apocalyptic apocalyptic world Where everything is overridden by jungle and so the civilization is no more. There is only Scraps there's only Broken stuff broken down stuff there's this Rotten thing that you have to look around but so that is so much life with The jungle as you can see here again, but still there is pretty heavy machinery going around seat and Beneath the first layer off of the ground so We have even small pockets of civilization like here where people just build houses with scraps and Fill them with whatever they could find so we had to Change our workflow to texture paint because it was much much easier to convey what we would do do then building all that stuff Was mainly the easy part because my experience comes to the Reasoning that you have to build every step knowing and planning ahead for the next step that you have to come around so For the second step 3d layout and animation. I was facing a Problem that I never faced before it was that I had to usually do all the work alone in this phase. I Nearly prepared all the 3d layouts for Cinderella the cat and I couldn't do it for this movie because if I only did that and it was massive because a Whole movie is still 85 minutes. It's more than 50 scenes and more than 300 shots So we had to Split it and I had to build some tools that could make me faster in doing my part and maybe even try to explain my workflow to someone else just to get along so Even for example our typical layout workflow is pretty simple You just have to take the set take the characters and link them into an empty file This is pretty easy. Usually it doesn't require that much explanation, but even though that is quite easy There are some recurrent layout issues if you can happen to not having named correctly your collection when you link it you might not Link your character at the word center and if you don't notice that problem Before you start animating then it is hard to extrapolate the same the same position if you didn't know Probably you won't display correctly all the armatures you the rig UI was not running It was all kinds of these issues that simply I was prepared to to face because I always did but Every every one of those means I had to lose time and every time I had to lose time the movie this time so I took a simple approach and broke down every single action I had to make to To go on in the movie and decided to Create just one button for every task. So if I have to link the characters I have to have one button that links the characters in the in the scene and controls if all those issues are already Looked upon so that I don't have to do it afterwards This comes with the bonus points that explaining to somebody to just press a button instead of How they would find the character and then important in the scene and then all many other small but For me it's simple, but not for everybody little tasks was much much better and I never code it before I'm not a developer, but still I had to to learn I try to find some Some way to understand how I could translate everything into an add-on internal for a studio and that was surprisingly Easy not because I am some special genius because I'm not I sure you're I'm not but because there is So much information aligned. There's a very big community. Thanks stock over flow. There is a very nice nice general approach to the Blinder community to just Share information on how they did something. So what I'm trying to do here is trying to share what I Found doing all this so that maybe Everyone can benefit from it So what I learned from this is that everyone can code because if I did it everybody can But not Anyone should do it because inside the production you have to pick correctly who ever chooses your your tasks so since we talked about the Linking phase with the 3d layout phase. We have to we can move along and talk about the rigging phase and I Would like to quote the director Alexander Rack of this movie that Told me I want to make a movie with just two characters lost in the jungle and nobody else. It was a Nice Information having a whole movie in which you have to just create two rigs It's quite it's quite a good news Then the story evolved and you might judge for yourself with this little Little montage I made if that's what happened if you can see you have two characters But then it happens There's another one because you cannot have just two characters inside a jungle that Interact with each other and then a whole bunch a lot of people and There are animals. I didn't think about animals because inside the jungle that cannot be just people they have to be Populated with something and so our Character count goes up it goes up and up and up If I remember correctly in the Cinderella cat we had roughly Consuming all the variants 200 and should show the 250 characters Here I can give you some numbers, but we went way overboard with it We started we do Characters and obviously the director lied to me and now we have 12 base Main characters of which we have 43 variants 90 secondary characters with 46 variants because there are many per se and 62 different animals with The exact same number of variants so this puts the total to more than 300 rigs and I was Basically alone in rigging because it was only me who's going to take charge of that so I had to even in this phase think again and Confirm myself with the fact that I couldn't do everything. So I had to break down the rigging phase. I had to Get all the people that would Wanted to help me in doing this and try to break down the phases in something that would comprehensible for them so that way they could actually do What was useful for me and for the movie because let's face it nobody likes rigging but anyone can help as Long as nothing breaks up and that that's the hard part because everything breaks up when you break it I never done a rig that works at the first try. So What we tried to do we tried to simplify this We had one body as I said here one body to roll them all But it is a lie because they're basically five bodies because It's harder to do it, but it sounded more cool with one. So I went with one We had one body to For every body type. So we had a fat fat person thin person tall person short person the men women and that came to around to five type five different types where the The character modelers could just pick one body and then start to model it Without changing the topology. This is the first important thing. I made just the first rig with the all the vetx weights and all the The best as I could do at the time Rig that I could make and by modeling without changing the topology you can see I picked five random characters We two of them are the are some of the protagonists and the other three are general the one in the center is Is just somebody on the on the background and the the last two ones are some secondary characters As you can see they all have the same topology That meant that I could recycle UVs if I use them for the characters. I could recycle vertex weights for the most part and That was a very simple operation that the 3d modelers could do without having to worry about About too too much stuff Then after modeling without changing topology, we made a first rig with basic metric both position. That means we just moved around the basic metric bones and generated the rig and that was already kind of ready to To make some lead 3d layouts Then at the last part We make a second pass where the rig is refined and enclosed correctly This means that we have way less operations overall It is so easy that anybody can learn how to do it It is animation ready at more early in an earlier stage than what we had to to understand before and I get to be involved only once that is the most important aspect of this because every time I get involved in this the movie gets lower so I had to add a little note a little footnote that's most of the times because I learned that However, you plan you may plan everything you want but still something might Not go the way you planned, but still it's good So I'm still trying to make you look at numbers I know hate me all you want, but I wanted to point out all the animals we made because doing animals is A really interesting part because if you have a used rigify rigify has some Pre-compiled metrics that you can use to build your personal rig It has some animal variants, but during the movie we had to make Many many more as you can see here. We had quadrupeds that mainly covered by rigify already birds That's hopefully covered by rigify many exapods or more And this is at all not covered by rigify some primates, but that's basically humans and some water creatures and we have only sharks I think in rigify actually right now and that is Not the same thing as a chetashian because if you know the Back fin goes different way so we made some Metrics that are in the process of getting adapted to the new rigify because Something changed along the way as you might have grasped from the 2.79 b and 2.80 Rigify changed a lot between these versions and it changed a lot again in the 2.9 and how in 3 and 3.3 so I had to We're in the process of Maybe updating these metrics so that they can be added in the next rigify distro so I have to talk right now about this What happened why we used 2.79 b while there was 2.80? To have this I have to give you some time of reference We started pre-production in july 2018 We had our Major workflow definition is if this presentation was the real time I would be nearly At this time talking to you because we had all our rigs defined. We had all our models. We had our Our workflow was predefined and we Start will start animation around february 2020 and Blender 2.8 is out on july 2019 It was almost a year after we started and that's That was pretty hard because There was all these magical fantastical stuff happening in blender 2.8. I wanted to implement everything I wanted everything then 2.8 could give me so I tried to make a little bit of screening I tried to To bring everything I did on the for the for workflow in blender 2.8 But I'm going about it in a minute on what I learned and what I what I see what I saw Just for reference. I had the this production end because we finished on july 2021 You might notice we took a bit more than was In the 2.5 years Time reference I gave you at the start. That's mainly because cobit then So I had this new fantastic 2.8 version And I wanted to use it It had a new collection system. There was grease pencil more useful than ever ev I I loved a tool like ev I will really want to use it There were new constraints new to viewport updated cycles overrides better performance overall It was all I could ask So why didn't why didn't I use it? that's because every one of these pros comes with a With account So our scripts heavily relied on display layers instead of that on collections and With great power comes great responsibility and collections are basically Display layers on steroids. So I could not change my scripts relatively easy We had already defined our workflow. So change it was Not so much in In the box for us We had to convert all our bundle render materials as much as I loved ev It was not compatible with blender random materials that we use for pre-visualization So I had to change everything We had to rewrite most scripts because The internal blender structure changed. I don't know if you If that is clear how much 2.8 was a complete game changer for the whole structural blender And that means even for people The for example, I'm just Giving a simple example the change from right click select to left. It was something that I had to Explain to every team member and not everybody is always Ready to these kind of changes you go my muscular memory you go by many Habits and I know you could switch again to right click select But it was such a major and important change that I thought it was Still better to keep Blender default settings 2.79 b was a two years old release. We'd already finished fixed major bugs. This is the most important On of the things I had to consider because 2.80 was stable. It was fantastic, but We didn't know what bugs it could come with using it while 2.79 b we used it for even for Cinderella the cat. So it was quite known for us what what to expect And as for overrides proxies seemed to be still more stable and I as I saw in the last version of blender I can finally use them, but up until 2.993 it wasn't they weren't that stable and Finally there was no going back option By going to 2.8 every file we saved in 2.8 was wasn't going to come back to 2.79 if something didn't work so it was Actually pretty hurtful for me. I had to choose why serve in heaven when I could just rule in hell and I'm not talking better. You blender 2.79 b. You're not hell. You were anything please So the importance of an lts version from a studio perspective I like to look at this in three Kind of perspectives Looking at it from up from the past you can experiment with new features Knowing you're not going to left behind if you're not ready if you have to change If you see that a new version doesn't support what you did until now you can Just stay in your long-term support version You can you can more easily find tools for your lts version since it's going to be stable for two years This is for whoever uses external add-ons the Everyone who writes an add-on has to maintain it and knowing that you can write an add-on and keep it for two years is it is actually More easier from a managing standpoint. So for a studio that needs to use those add-ons It is actually really really important knowing that you can You can afford to use that add-on and it is supported at least for Two years for your version of blender and you can plan on developing your own tools If you're a studio and you need to develop something you will know that it will be used For example, I had to develop many tools that are not any more compatible with blender 2.8 But we're going about it so With this third movie with blender 2.80 out of the question we had to Look at ourselves and think how can we Go on. How can we better our distinctive 2d look? What changed during these years we started with the art of happiness That is a full 2d theater film with 3d elements in it I was in I worked on every 3d element you see in the in that movie that it's not what we're talking about now but It is mainly 2d characters. That is the most important part that you have to to understand for the difference Those were hand-drawn characters moving in a 2d space But we used 3d elements here and there for example in this image of the little kid You can see the the grass on the front and The tree on the back are both 3d elements while the kid and the the sky are not In Cinderella the cat we had 3d characters layered on 2d paintovers because we try to keep our post-processing process that we used for the art of happiness and brought it in Cinderella the cat Where we used our 2d characters like in disney's vertical camera and but we used the 3d backgrounds To let the 3d characters interact in a 3d space And then we rendered those 3d backgrounds and had our background artists paint over them to have a 2d layer to use in the compositor In Jaya and Lenny the work in liberty. It is full 3d CGI. We have Not made this paintover effect. So everything you see was painted over but in form of a texture on the On the background and then used in post-production to get our distinctive look I'll make some example here as you can see in the art of happiness. We have Even I like to point there's even more of a change in our stories because in the art of happiness it is the story of Sergio Taxi driver Naples and so the whole movie is basically Two or three guys inside a taxi. Everybody is seated down You have very For the most part the same Same shots inside this Taxi there was a 3d taxi where we painted over but the characters are 2d and aside from the the background The movie background that is 3d and some elements like the steering wheel. It's all 2d Then we moved on to Cinderella the cat where everything you see here Is a 3d model that was exported in layers and then painted over This is A really big change from the look of the previous film. I think everybody can see it But the big change for us was that now we had 3d characters We had to make them interact with real things. We cannot use hand drum anymore So they had to be moved inside a 3d space and That was the main The main use we had for Our 2d backgrounds You can see here all the characters have this rim light on them. That's Not 3d generated it is It is used without the process of composing that gives it a more of a 2d look Here in Jaria Lenny you can see everything is 3d, but we are heavy compositing and Post-processing everything still has this 2d look about it All the rim lights you see here are used with the same technique from Cinderella the cat so that we can keep that The same feeling even though now it's everything on in 3d Okay Just to make you understand I made this little Video that makes Pretty clear it is just a visualization. It's not actually stacked like this But you can see the the characters are on two separate layers And the background is generally divided into even more layers Just so that you can in post-processing then layer them one on top of each other and have With using various passes this final result that you see here so uh, our way of doing this is like splitting the atom so you have to take the The final frame and understand that that final frame is the result of combining Many of these layers. This is just another visualization of what you've seen before But it is important to understand that I had to prepare A system that took into account that we had to export everything separated like this And it should not hamper the animators work while we try to make all this system And we didn't have a fixed amount of characters on screen. We had To create this system having to take into consideration even more than two characters because as I said before The director lied to me. There were not just two people in the jungle So I made this in the this system that takes up to six characters And every single one of these layers, for example, I I took here the the background layer Is uh, split up in different render render passes that are then used inside the compositor So there's a color pass. There's a dark light pass There's a the ambient occlusion. These are the faked 2d rim light and a mist pass Um, there is Basically all the lighting we had done Is just in the direct light pass As you can see we use, uh, 3d lights just to get the general direction of light so that the Composer could just Use it as a reference to make their fake 2d lights Using the mist pass so that they that we use it basically as um A depth pass because we use the z depth In the previous productions, but the the problem we have With that is that it is parametric So it changes the the start at the in the last value In depending on what's on screen while using the mist pass We could just decide to have a gradient that goes from the start of the camera to how much far we want it to to consider And you can see here I had to be sure that all this system that breaks everything down and combines it all over again Can Doesn't break during the animation phase because we had to have a pre-visualization. There was Already Near what the director had in mind so that he could so that he could give some directions Still director or nonetheless About the light principally You can see here in this previous that goes to the final That all the highlights on the plans Are the same even in the final In the final version of the shot. That's because we had Even in pre-visualization this the power to To notice that So I had to build specific tools because as I said before I couldn't do this all on my own. I had to do it on Cinderella the cat and I But I was in a different position. So here I Would not be able to do everything alone. So I had to At least try to Make myself faster. So I built some tools. I built an old tree updater and A tool to match the lighting between blender render that we use for pre-visualization and cycles that we use to to experts some black magic I like to call them like that some black magic buttons that the Somebody else would not Understand what did but did what they needed and some internal render Submitter those are some tools I had to build so that everybody could Set up their files. I made An example here In this scene we have The ability to change the direction of light. This is the pre-visualization scene In every file there were a pre-visualization scene A character export scene a background export scene and two scenes for the Masks that we needed to output for the background and for the characters Because when you transform everything in a 2d layer you have to Mask every character with Every interaction it has if it's going to be on top with the background that is on the bottom So you can see here the animator could already change the Delighting on the scene and when that goes to render you could go even in In the character scene and isolate the Lights because sometimes when you're in the jungle Not every light comes across between the The trees so there were many instances where our characters weren't lit enough So we had to make this system To separate the lights of the characters We have many shots where the background has one direction of light and the characters have a different one But obviously compatible with that otherwise you would have noticed in a bad way And for example here you can see Someone is sending up lights and and This is I I can say I messed up the title of this slide, but it's the same Here I'm making you look at how this tool works For example, I opened this file and it was all broken because you can see there is only the character inside the An empty void and it is not actually what you have to output. So I made just The famous one task one button button that updated the file to the latest version of the Output nodes and Set up again all the background and all the levels as they should be Then you can go for the export part in the background scene I'm making you see here that what happens in the node editor, but all this This add-on was made in with the first principle that Nobody should go to the node view Just because everything every time you change something in nodes Nobody does nobody knows what happened It is too hard to go there and try again to understand that and we had more than 300 shots So it was unthinkable that you had to do this for every single shot So with that system on the left You can see I could Just switch on or off different levels of the background and so that they will get exported or not And I could set up if that was going to be Using the the same lights from the characters or not. There is even another little gimmick I made here Sometimes we had Scene square the camera was not moving at all So if the background is not moving and the characters are acting at the center when they don't don't move Don't move that much It was useless for us to export all the sequence from the background So we just exported the first frame and then layered the characters on it So we had this the this little button that could just render one frame And but it memorizes the previous duration in case somebody messes up You can see here I made the same thing for the characters, but the node view is a little bit more complex because There are as I said before Six characters so we had to take into account that every one of them had to Have their own personal output nodes and they had to be switched on or off I had to have a comprehensible interface because I I was not going to check if If a particular angle in the in the node header was on or off I needed something that was as easy to look at as this one So you can see I just want to export these two characters and I want to set up even the Fake rim light we had three presets to use as in near the camera in mid-ground and in the more far from the camera all these tools that I used were really made for for production, but As I said before I really wanted to use 2.80 So it's Even though the movie ended more than one year ago I I still think how could I do it today? How would I do it today? and First of all, I would use collections instead of display layers because those are Much more powerful as I said before with they are so powerful that it is For this particular setup It is not that good. So I have to Start to understand the way to minimize the The options because there are too many Then I would use evie for pre-visualization and masks and cycles for rendering This is Pretty big catch because evie for pre-visualization is much much more faster than blender render So it would have been much better if we could use it I would make it all asset manager compatible because building a whole jungle was big they had to make roughly more than 70 different kinds of plants we had to make Tense Tense of different assets and so having a nice way Of looking at them with a preview and importing them the scene would have been Huge a huge way to speed up our set set process I would use geometry nodes instead of particles for vegetation because every time you saw the ground it had some particle particles for grass and having the The chance to use geometry nodes in our newest production. I will totally use them round instead of particles And risk pencil for special effects As I said before we have Really heavily oriented on compositing production but risk pencil offers the the chance to use All this all this stuff directly inside the shot and it is huge for us because Everything that gets Moved on through the production line is just something that you have to lose more time over it Going on so all the things you can do Before they become a problem are perfect so Thank you for watching I hope it was entertaining and I don't know how much time I have So I'm gonna look at it It's nearly five minutes. So if you have some questions, I'm here If you don't so say Yes Thank you for that question. It is As I said before in the past we use the z pass to the the z depth pass to to make it it is actually A pretty crude You just take the the z pass and overlap it with itself and tilt shift it with Two or three pixels and then you just you have to do more than that But this is the the core process as I said before here we use the mist pass instead of the z pass because we had every time to set the start and the end of that Z depth pass To have an effective rim light And the bonus point is that you don't have to put lights Around the scene to have the correct rim light, but you can just export it We tended to export two rim lights one for always without even having the Without taking into account the lighting on the scene We always exported one from one side and one from the other side And since it is a 2d effect If the camera was tilted it was still going to go Like the the image so that's basically the the gist of it, okay If nobody else needs to know anything