Fluid Sims and the Future of VFX

This has been something on my mind lately, and I’d be curious to hear what other FX artists think.

It feels like it’s only a matter of time before we’re going to be able to run volumetric fluid simulations in realtime. Nvidia’s already doing a lot of work getting volumetrics to run in games like Arkham City or Assassin’s Creed:

I’m curious what you guys as FX artists would think of a shift like this. Now while I doubt traditional sprite-based FX would be completely replaced, I’d personally be a little disappointed if volumetrics became the new norm.

What I love about sprite-based VFX is that the limitations force you to get creative. It’s fascinating seeing how many different ways FX artists choose to represent things like fire and smoke, the creative techniques they use to give it their own look and motion. You get such variety in the artistry.
But when you look at fluid sims from different people, a lot of them tend to look very similar, because so much of what you’re seeing is being created by the algorithm rather than the artist.

Of course I still feel like realtime fluid sims could have their place in how we do VFX, it’d just be an over-reliance on them that worries me.

What do you guys think?

Respectfully disagree. Whether you’re doing offline or online simuations, a large amount of artistic direction and technical expertise will be required. Sure, Houdini lets you create an explosion with the click of a button, but no one in their right mind would let you ship with that instant result in neither games nor film. I think you should take the time to break down various VFX-heavy shots more closely, I think they deserve more credit.

As for the realtime aspect, while I agree that there’s enjoyment to be found in working within limitations, we’re also stuck constantly trying to break out of said limitations. Having more robust volumetric tools would be just another tool in the toolbox, as limited as any other, but it would help us push our craft to another level. I heartily recommend this presentation by Sébastian Hillare on his research into volumetric rendering in games - it provides quite a number of examples on how these techniques could be used to great effect.

5 Likes

I’m with Chris on this one. Game fluids might look similar because it’s a newborn tech with incredibly limited functionality.
Are you really saying that actual, rendered fluids are looking too similar too? No, Fume, Houdinis shelftools and other 1 click solutions obviously does not count.

Also, a widespread use of realtime fluidsims is many years out. Not something we need to worry about anytime soon.

I suppose I’m being harsh, I didn’t mean to sound like there wasn’t a lot of skill and creativity put into using fluid sims. Rather what I meant is that there still very often tends to be a fluid sim “look”. There’s an aesthetic to them that is easily recognizable. Simmed effects tend to look too smooth and flowing, without enough rough granularity you see in real smoke and fire. And I’m not just talking about stuff you see in games, film quality effects have the same kind of aesthetic.

I figure if it’s going to look fake regardless, I’d rather it not all look fake in the same way.

Like I said, I do feel like volumetrics will have their place, and I’ll be interested in seeing how the realtime VFX industry adapts to it. I’ll definitely give that article a read, sounds interesting.

Interesting opinion. Would you care to share some film examples?

I wish I could find higher quality video of this one. When I saw the explosion of the thermal oscillator in The Force Awakens, my first reaction was how ‘buttery’ the explosion felt. It was just so smooth and bulbous, not nearly enough violent forms in the fire and smoke.

This one from The Girl with the Dragon Tattoo was cited by Freddie Wong as being almost indistinguishable from real life, and while I feel they did a great job replicating the color palette, the movement of the fire again feels too smooth, and this time kinda ‘slimy’.

This one from Zero Dark Thirty is interesting because it cuts from a practical explosion to a fluid sim, and the difference is jarring. The sharp violence and color variance of the practical effect is contrasted with the smooth sameness of the CGI version.

The reason I worry about an over reliance on fluid sims in realtime VFX is because there’s already an over reliance on them in film. My personal view when it comes to film is that fluid sims should only be used in circumstances where you could not otherwise do the effect either practically, or by comping in practical footage. Granted I understand it’s faster, easier, and more cost effective to do effects digitally, but I’m just speaking from a purely artistic standpoint.

I was disappointed that The Force Awakens relied so heavily on fluid sims when the prequel trilogy did some of the best compositing work I’d seen for effects.

This is why I like games that utilize practical footage in their VFX. You get a lot of those rich details that you just don’t get with fluid sims.

The recent Mad Max game combined fluid simmed smoke with gorgeous practical fire elements:


And the Call of Duty series has had some very pretty effects thanks in part to practical footage:

Cool! I don’t agree, but it’s nice to see how others think.

I’m way excited for more real time fluid sims, personally. Real time fluid sims are just another tool in our toolbox. It’s not going to replace us as artists, but it will allow us a wider range of tools to work with. I think real time fluid sims will help push our medium forward.

Also WOW Arkam Knight is a beautiful game!

4 Likes

As i see it. Simulation by the core of it to simulate realistic phenomenas . so when you want to create realistic look and not much stylized you will work with sims. When the art direction doesnt request that realistic look so need for that “tool”.
Other Point of view in my opinion, each artist manipulate the sim in different way to achieve unique resault or behavior. Otherwise the way you see it is kind of canceling the reason to have the big request for vfx artist in films, if its just a matters of algorithm that solve it all.
All in all intresting topic.

I used to have to make effects to fake volumetric light cones… now the engine pretty much handles it. I’m ok with that. Now I have more time to focus on other FX. If the real time sim tools get close to our offline tools are currently, you are still going to need an effects artist to be dealing with them.

For me what it really comes down to is this… Our tools have and will contiune to change over time, but the need to communicate game play to the player has not.

nick

2 Likes

Love this discussion. I agree that we’re a short ways away from having practical usages of volumetrics and fluids in realtime, paralleling the progression of films. I fully believe that our workflows and broad concepts are about 10 years behind, but closely following the capabilities of, film. Every time there is a new console jump, the entire real-time industry jumps to say “Hooray fluids!!” but they always skip the early steps that film required to get there. We’re currently in the middle of a massive technological innovation state for games, but not a lot of people recognize it:

Shaders

There was an era across the early to mid 2000s when fluid simiulations were still far too costly to use across full productions. Think the Balrog in Lord of The Rings: All of the fire was still done on sprites, with amazing shaders. The tools of the era were all about putting super complex and beautiful shaders on basic sprite particles (oh man it’d driving me crazy that I can’t think of the names of these software… edit: AFTERBURN!)

In games, we can do AMAZING things with intelligent shaders, vertex shaders, compute shaders, etc. We can handle complex lighting on particles, beautiful procedural or physical motion, stunning scene integrations, and even player interactions within our shaders. If shader toy is an example of anything, it’s that marvelous things can be made with shaders. I’ll also say that anyone who can correctly use shaders in this industry has incredible job security.

There’s another component of this that I find matters far more than we give it credit for - our authoring tools usually aren’t keeping up, and most companies don’t prioritize these improvements. Our team at Naughty Dog was extremely unique in that we had a dedicated graphics programmer for the better part of 5 years (starting tools from scratch though), and now I’d argue that they have some of the most amazing in-game vfx interactions in our industry. Most studios don’t have this, and our only saving grace is being able to point to the amazing work that the team at epic is doing to increase our potential, or external tool integrations like those from the great team at SideFX.

I’m also seeing a massive counter trend away from realism in FX towards stylized effects. Largely, I view this as a direct result of how technical our tools and workflows have become, pushing a lot of very artistically talented FX artists to find beautiful and elegant ways to make their FX captivating without the needs of insane math and abstract workflows. They probably get to go home from work a lot earlier for it too. For new artists entering the industry, there are far less external roadblocks (and thanks to people like @Keyserito far more resources) to become a great stylized FX artist.

All of this rambling is to say that while I definitely agree we’re moving towards real time fluid simulations, we’re in an earlier stage of technology where we should be focusing on the same exploits that happened in film - shaders and integration. I contend that we won’t widely get fluids until we step through this process and build a more evenly balanced (art vs tech) workflow to utilize them.

On a very personal level, this might be the largest reason I’ve been pushing to make this forum with all of you - so that we as an industry of artists can share more of our techniques and develop our next year to ten years of workflows and expectations together, and to allow people outside of vfx to get a glimpse of the world we are living in.

13 Likes

Just another tool in the toolbox. If it’s appropriate for the art direction of the game and fits within the GPU budget then why not use it? I’ve used pre-rendered volumetrics for in-game effects and cutscenes so it won’t be much of a stretch to use real-time volumetrics.

I’m not really fixated on sprites and particles. I write custom scripts to generate meshes for smoke, trails, etc. and write custom shaders for all kinds of odd effects.

yeah that what I mean with my point. the sim itself is just a tool to help the FX guy make what ever needed and even be creatively used in different use. what I tried to say that its not a math magic that will take the FX jobs from the people :wink:

Edit: I also believe shaders is the future and real time rendering feels like it can do better then offline (if the tech behind is really advanced) that why I also find this RTVFX Role best for me because its the perfect bridge of technology and art (not many people out there know what it really require to be one). I’m not sure if just a great texture painter will be able to implement a full working in game flowing effect and vise versa for a full shader programmer :slight_smile:

God yes! If tools could catch up to tech at this point, so we would be free to use the power available without jumping through hoops, there will be innovation abound.

1 Like

I agree with others that it really comes down to the art style. I much prefer stylized vfx. Complex shaders, morph targets, and flash animation are my bread and butter. However I’d love to be able to do everything in real time; the more tools the better!

I agree with TheVman… Not most case but many cases there are all similar explosions in movies.
Making explosion textures and blend with layers makes game’s style. I think it is very important.
I’ve read post about this subject http://valdism.blogspot.kr/2014/02/fluid-simulators-and-future-of-vfx.html

That’s actually my blog lol

This post was kind of a reinterpretation of that blog entry. I was curious to hear what the general consensus was from other FX artists.

1 Like

I would argue that as photography has not replaced painting, sims wont replace artistry. Tbh, as realtime fluids sims will become more accessible in a pipeline for game applications, people will get creative with it just how we got creative with facing quads over the last decade.

To be honest, I think the “all explosions look the same across different movies” has less to do with the artist being bound by rigid tools, and a bit more about harsh deadlines, changing shots without changing vfx production timelines, and the general shitstorm VFX studio find themselves into on holywood movie productions.

1 Like

Keith, I couldn’t agree more.

What has gotten me more interested than real time simulations so far (and occupying most of my free time outside of work… yeah I’m a nerd…) has been raymarching procedural noise in volumes (a mesh cube) and using those as particles… Makes for some really interesting results, especially when it comes to smoke and density based rendering.

1 Like

You should have some conversations with @mattorialist - he was pushing approaches like this quite hard during early production of Uncharted 4, but had to get grounded by efficiency needs during the production. Still, he nailed some amazing looks that I think are extremely promising for the coming years.