There is one thing in my VFX pipeline that i absolutly hate: the fluid sims.
While they do tend to yield the best results for complex smoke movement, especially the “pyroclastic” one, it takes me dozens of iterations and sometimes - months of work to generate a single decent flipbook: https://www.youtube.com/watch?v=vfB3HgaYjmA
That’s why i tend to evade flipbooks as much as possible and only use them as the “last resort” measure. There are other reasons why flipbook usage might be unpreferable, such as memory consuption, repetition, e.t.c., which are all solvable to some extent, by my biggest problem is the iteration time.
I don’t seem to be getting better at the fluid-sim craft with time either: it still takes a lot of time.
I’m not sure if it the consequence of the tools that i’m using (maybe blender 3D is not the best choice for fluid-sims), my approach or just a lack of knowlege.
What is your experince with fluid simulations? What software do you use and how long does it usually take to get the results you need?
Using any high end software for sims (I use Maya or Houdini but am new to Unreal, and realtime fx) it will take a while to learn the tricks, but once you do you can get decent sims out in a few hours or a day. Just takes the patience of learning for weeks, months, etc. Still it will never be as quick as Embergen.
Thx, i did chech EmberGen out, it looks promising)
But it’s a relatively new tool.
You’re saying that you are new to realtime VFX. Have you ever created a flipbook for a relatime game?
I don’t have any experirnce with offline VFX, but i assume there are some corners to be cut, so i.e., the simulation doesn’t have to look good from any side and work in every lighting coditions, it’s possible to adapt it for a specific frame.
Is 2h-1d the time it takes you to create a flipbook for a universally-usable real-time VFX or a cinematic single-use one?
I would pick up Houdini. A lot to learn but you have as much control as you want or need. It takes a lot of learning but is worth it when you get there.
There are good export nodes in Houdini labs, like the texture sheets ROP. The texture sheets ROP can automatically render out normals from a light rig (which is a pain to do manually) and will build your flipbook for you so you can really streamline your workflow.
You can also use wedging to generate multiple texture sheets by changing a seed value in your system.
If you want to learn pyro in Houdini, check out Steven Knipping’s Applied Houdini course. Really worth the money.
Yes, you cut corners but thats not the reason high end sims take so long. Its because you really need to learn the tricks (tons of tutorials for Houdini) to get the good sims faster. And then yes of course it depends on how fast you PC is and how high res your sims are. For games they can be lower quality, depending on what you want so the actual final sims or renders can be quick. But again nothing for film is realtime like EmberGen and while still new, its great for many types of sims and good for production now.
As mentioned Houdini is best, but you will have to put in the long time to learn and use it. Also its great for Rigids and all the SideFX Labs tools to use with Niagara, etc…
Damn! You might be the fastest artist I’ve ever come across if it only takes you that long to make a production worth flipbook.
For me, nailing a good smoke flipbook along with the shader to go with it, testing it, iterating on it in all the different usecases takes at least a few weeks. Making a good intial sim is the first 3% of the work for me. It’s all the work that goes into proofing it for production that takes time. Hell, even just testing various density levels to see what gives the most mileage. I mean, most games have what, 5 smoke flips, tops? Making sure that the new one fits in the suite and covers the most possible usage is tricky.
That said, I’d estimate a few weeks initially, then I’d spend another one a few months later when it’s been used and abused for a while and you know what you need to fix.
Also, Embergen ready for production O.o… Can’t say I agree with that. I’ve seen one production example so far and that had… issues. Very promising piece of kit, but it’s far from ready in my book.
Hah, no no,I am brand spanking new to realtime fx so please take anything about “in a game” with a grain of salt. Sorry, I was speaking from film experience.
And of course I didnt mean final when I said decent within a day. I meant that after taking weeks and months, even years to learn the ins and outs, through simming and rendering forever locally and on farms, that you eventually learn how to get a decent sim within a few hours or days instead of getting frustrated when it looks crappy after an hour.
I realized something from speaking to a few of you guys. You are use to realtime where everything is instant, and may have trouble being patient to take the longr times to learn those tricks. I have to tell you, its worth it, if you have passion and love it.
Also you have so many tutorials and helpful people on forums, so @LudeosShaft learn learn learn!
I came to realtime vfx from a career in film/commercial work where I was almost exclusively specialized in simulation and rendering. It does take a long time to get the feel of settings and cut down on iterations for sure, but I agree that the years of experience add up and you do get faster. For me, Houdini has been a life saver. I always start my sims from a previous project where my settings are already pretty dialed and modify from there. I also reworked the texture sheets tool that comes with SideFX Labs (Previously game development toolset) to use Arnold instead of mantra which started saving me hundreds of hours in render time. I haven’t checked out the new render engine Karma but that may replace my hacky Arnold solution. With my current setup I can get some quality fire, smoke, or explosions in 2-5 days depending on the project and client feedback. I work for an outsourcing company so we have to be pretty agile and we work within every budget, quality, style and platform restriction imaginable.
I will say, I started the vfx department at my company and training my artists in simulation and rendering is always the final stage. It’s such a deep field and as we’ve been talking about it takes time to start seeing satisfying results. So I get everybody solid on shader math, laying out uv’s for vfx animation, making textures, and all those aspects of realtime vfx before I ever plunge them into simulation. When I do I give them all my pipeline tools and my guidance and it still takes a good month before they’re making anything quality on their own. A big hangup a lot of my guys have in the beginning is keeping the resolution of your sim as low as is useful instead of waiting for a full res sim when you’re still making big decisions.
I second what @HowardM says, just keep absorbing tutorials and stick with it. For real the last 8 years of my life have been dedicated to simulation and I’m still finding ways to get better and faster. I did my first fluid sim on a garbage laptop in college. It took 2 weeks to sim, 2 weeks to render, and it still looked bad haha.
Thx everyone!
I did try to install Houdini las night - but it turns out that it doesn’t support my old Win7. Guess i’ll have to pass on it until the inevitable upgrade happens)
As for years of practice - i don’t have much of them: i only switched to photoreal VFX from stylized ones, like, 2 years ago. It both scares and encourages me that once i approach the 8 year mark things will get better) I do believe i have the knowlege to get an acceptable sim going, but most of my time goes towards what @Partikel said: iterating upon the sim + all the other stuff in the pipeline to make sure that it looks good in game.
There is also Fluid NInja VFX tools for UE4. I bought it recently but forgot to use it. It’s a 2D fluid sim but it looks very promising:Fluid ninja video
