Kangjoo Kang followed his passion for crowds since he was at university. Now a crowd artist at eNgine Visual Wave, he has worked on various epic productions in the best Korean studios. He shares his story with us today.

How did you become a crowd artist?
I majored in CG at the university. My brother, who was studying with me, showed me a video: it was "World War Z". I first learned about crowd simulation thanks to the scenes in this movie. It really stroke me, that's how I started studying about crowds.
At that time, I knew nothing about crowds or crowds software. I had a hard time getting to know the crowd software Miarmy, installed it, and used it. I had no access to a render farm as a student, so I worked all night installing Miarmy on all computers in the school lecture room in order to build a render farm.

I made a lot of efforts to be a Crowd artist. The most memorable moment was when I thought: "Ah! I'll be the agent!". So I played a passerby in a movie and had to follow the instructions of the director. This experience helped me understand how a crowd should move. What I felt at that time was that if I controlled the movement of the crowd well, I could make a small number of people look like a very large crowd.

I started working as a Crowd Artist at Dexter Studio. I was lucky because my team leader at that time gave me a lot of time to work on R&D, so I was able to do as much as I wanted. I was also writing my master thesis, comparing three different crowd software so I was able to do a more detailed analysis. What I focused on was the speed of work, ease of modification, and the quality of the simulation itself.

That's when I studied Golaem in details, and discovered the Golaem Layout tool, which I thought was a powerful tool to get the job done in a time a constant updates because it was enabling to perform modifications after simulation.

Could you share a few projects you have worked on?
I am more suited to crowd work that expresses creativity such as jumping and swimming than to human-type crowds. It's a very attractive task to control characters movements and their interactions.

I put that kind of work in the first scene of my showreel. It was a very interesting and memorable project where fish attacked an individual.

After Dexter, I moved to Digital Idea where, among other productions, I worked with Golaem on the movie "Vanguard". On one of the tasks we had to do, the FX team was saturated and the animation team was not able to work with a large number of objects.
I think Golaem's Physics function is one of the really good features so I thought: "What if you use the Physics function to perform rigid simulation, remove the cache, and make additional corrections with the layout tool?". I worked on that hypothesis, and the results were very surprising.

Given your experience in crowds, which trends have you seen emerge over time, and how do you see the future for crowds?
The location of the crowd in a shot keeps evolving: the crowd used to be behind main characters and play an auxiliary role, however now it is getting closer to the main character.

I think one of the reasons is because higher quality animation sources are available. Crowd software are also getting lighter and lighter and their simulations can be exported to a variety of platforms, including Unity, Unreal Engine, and Houdini.

I think that if crowd simulation has only been used in the past and present to make movies, it will be widely used in the future in the game industry and in research that many others need. :)

How did the Covid crisis affect film production and the way you are going to work with crowds?
The Covid situation seems to have made the role of crowds more important.
It seems that there will be more opportunities to add CG crowds to a scene because now we cannot have many people on set.

I hope that Covid disappears soon, but it has become an ironic situation in which there are more opportunities for crowd work.

Would you recommend CG artists to get in crowds? How? Which skills should they develop?
Of course! It provides artists with work efficiency and includes workflow in many domains, including animation, FX, Simulation, etc.
I am still continuously studying various fields: animation, motion capture, and script... A Crowd TDs should know almost every part of 3D.
If you want to gain a general experience, I think a good way to do it is to work on Crowd Simulation.

One thing I felet while writing my bachelor's thesis on Crowd Simulation was that I felt the program was not so important, and the most important thing was my own feelings. If the picture I want to express is clear, then I can use any program for expressing the picture.
So when I recommend Crowds, I suggest you analyze how the crowd works and think about how it works, rather than how you use the program!

Would you have some advice for people wanting to create a demoreel in order to be hired as crowd td?
Crowd TD demoreels should not be just large numbers and huge shots, but also high-quality work on crowd movements and interactions.
I think it increases the probability of employment.

Indeed, recently the trend of crowd work has been to continuously come closer and closer to the camera, and even a small number of characters can be handled to the crowd department. I think detailed expression is the most important even if the number is relatively small.

Anything else you would like to share?
My motto in crowd work is 'Continuous Hypothesis and Proof'. Crowd TDs constantly form hypotheses, prove them, and work on them.

I think Golaem was the best way to realize my idea. My dream is not to be satisfied with my current position, but to work harder and become a Crowd TD model for many people. I hope we can share a lot of information and enjoy working on crowds.

Dan Warom is a CG Supervisor with a long experience as Crowd TD. He has work in some of the best studios where he implemented crowd pipelines allowing him and his teams to create tons of crowd shots with an incredible productivity. Today, he shares his story and experience with us.

How did you become a crowd artist?
While I’d studied animation, I actually got my start in features as an FX TD. I was working at a company called Asylum when I was asked whether I had any Massive knowledge as that was the next project available to me. I figured how hard could it be so I replied “I’ll give it a shot!”, a paycheck is a paycheck.  So I guess I got into crowds because of cold hard cash!  As it happened, after my first Massive project, the main crowd TD left and it was just me and very much a trial by fire to get more skilled.  From that point on I was able to merge my animation training along with my FXTD experience to become a better crowd TD. Some 14 years later and here we are!

Could you share a few projects you have worked on? (if not mentioned in previous question)
There have been a few for sure, I have been very fortunate to have had the opportunity to work with some great teams on some fantastic projects but the few that stand out to me are; 300 : Rise of An Empire, simply because this was my first supervisory role where I was able to help build the crowd pipeline my way.  World War Z because well, Zombies and experiencing a completely new way of approaching crowds through MPC’s ALICE. Kung Pu Panda 2 was an incredible opportunity to develop agents that had far more complexity than previous builds so they could be both procedural or “keyframed” in Massive. Finally I’d have to say that last years Run The Jewels music video was a genuinely fun project to be involved in, from concept through to execution with Golaem crowds playing a fundamental role in the aesthetic.

Given your experience in crowds, which trends have you seen emerge over time, and how do you see the future for crowds?
Well I’ll always have a soft spot for Massive, it truly is an incredible piece of software that achieved so much and its sad that over the years its userbase has dwindled to a trickle. But that has very much been born out of a re-evaluation of what crowd tools need to accommodate in our industry. We need less standalone software, less black boxes, more extendability and better cross renderer support, something that almost all the big crowd tools now have. USD has taken flight and its brilliant to see Golaem and others support that natively now, meaning our pipelines can be a little more nimble in how we organize and render our simulation data.  I think we’ll of course see far more ML influence on crowd simulation, motion capture cleanup and retargeting and even just data capture itself, some of the work that Ubisoft has been doing in this space has yielded incredible results. 

How did the Covid crisis affect film production and the way you are going to work with crowds?
I think there was a initial global impact of all production going on hold.  We saw a lot of studio’s explore new avenues for crowds in the realtime and broadcast space which yielded a variety of results.  I think as a discipline in realtime there’s still some work to do which is why ultimately those results were only used sparingly but it was exciting to see how the realtime crowd solutions developed.  Now we’re seeing more productions go live and with social distancing and rules to protect talent, there is a renewed interest in effective, photoreal crowd solutions in all areas of our industry.  Fortunately with a wealth of training materials and tools to choose from, being a crowd td is a pretty exciting role to be in these days!

Would you recommend CG artists to get in crowds? How? Which skills should they develop?
Its really a question of whether you aim to be X artist with crowd skills or a crowd TD since they’re both very different disciplines.  In my experience, a great crowd td is a generalist who has a robust set of skills in all cg disciplines from modelling to comp.  There’s optimisations at every step of the process that you’ll need to get the best looking crowds in the most efficient manner. Even with renderers being as fast as they are, you still need to have optimized crowd assets to get the best iterative results.  At the very least, you need to have a base understanding of animation, motion capture and ultimately crowd tools.  Animation and motion capture because you need motion to drive your characters and the crowd tools to complete the simulation.  Tertiary skills like rigid body simulation for ragdolls, cloth simulation and complex navigational logic will come with experience but most software contains initial tools to get you started.  Familiarise yourself with Motion Editing, you can save a lot of simulation time by creating the vignettes you need manually rather than trying to chain actions together in crowd tools.  Crowds for simulation, always – I learned that the hard way with a restaurant full of talking animals.

Would you have some advice for people wanting to create a demoreel in order to be hired as crowd td?
First of all, remember the whole reason to have a demo reel is to showcase your best work. Not all your work.  So keep it brief and containing your best shots.  There’s a bunch of typical biped crowd scenario’s you can pick from, the classic army face off, the army charge, stadiums, city life or flocking.  If your doing large scenes, make sure to include some ground level camera’s, you want to showcase individual behavior as well as large group behavior, these are crowds of “things” after all.  If you’re a generalist, it might be good to include some turntables of crowd characters you have been involved in developing, whether that be one step of the asset pipeline to all of it.  Research tools offered by TD’s or schools like CAVE Academy, they have lookdev templates that will enhance the look of your asset turntables, a Macbeth chart and shiny balls go a long way.  Run turntables of interesting motion clips you have created and show them in context with a crowd simulation. For the more technically oriented you can take your cues from rigging showreels, if you have developed some cool looking logic that reads better in viewport w/ paths and navigation visible, then show it, a showreel breakdown can provide context to this stuff. Finally, something I’ve done for a while now is put a brief description either of role or work done at the bottom of frame, it helps people understand what your contribution was.

Learn more

• Read Dan's interview for The Mill about The art of creating a believable CG crowd

Want to become a Crowd Artist like Dan? Download Golaem PLE & start training yourself with our Youtube Channel