There's really no “fudge factors” in Houdini FLIP fluids, as you might be used to with Realflow. You should be aiming to simulate all fluid scenes at their true, realistic metric scale, otherwise gravity and other forces simply won't add up properly. If the nature of the sim feels off to you, then it's likely a problem elsewhere, not because large scale sims inherantly look better somehow.
I mean, large scale sims *do* tend to be more forgiving, by their very nature of being relatively slow moving compared to their grid size/particle separation, but there's really nothing about that look which you can translate to a small sim without just turning it into something it isn't.
A good starting point is to consider this: At a smaller scale, gravity doesn't change, so it will naturally be accelerating fluids across far more grid cells per frame, due to the grid cells being smaller. FLIP breaks down if you try to have it travel through too many grid cells in a single timestep… so typically even just for a glass of water sloshing around, you'll need at least 3-4 substeps set on the FLIP Solver… where a larger sim can happily get away with 1 in most cases. For an animating glass, I've had to push it up to 12-16 substeps to keep it under control.
You should also try to avoid low particle counts in general. You may think a smaller scale fluid will get away with less particles than a large one, but you're usually viewing it from a much more critical point of view too, so you need to push the particle separation right down until you're using at least a few hundred-thousand particles, if not a couple of million. The more the better really… it's a trade off against solve speed obviously, but in general, you can never have too much resolution in a fluid sim… certainly not within the tight bounds of current PC hardware. Max the resolution as much as possible within your project's time constraints.
Found 48 posts.
Search results Show results as topic list.
Houdini Indie and Apprentice » Flip Sim Scale
- danwood82
- 48 posts
- Offline
Technical Discussion » How buggy Houdini is.
- danwood82
- 48 posts
- Offline
The UI issues can certainly be a problem sometimes I'd admit (although a LOT of them are very much improved in Houdini 14)
But when it comes to nodes not functioning correctly - well, if Houdini has taught me one thing over the years, it's that if you think there's a bug in functionality, check twice, check three times, and check again… because it almost always turns out to be something stupid I did. Really, compared to every other 3D package I've used, I've never come across one where the core logic behind everything was as consistently sound and dependable as in Houdini.
It's a complicated package. It pretty much allows you to do anything imaginable - more akin to a programming language like C++ than to a conventional 3D package. But likewise it would be like saying C++ was buggy and broken. C++ *can* yield a buggy and broken program, but usually only because of flaws in the implementation you've come up with.
I know there are bugs, I don't presume Houdini to be perfect, but seriously, always triple-check your own methodology first. I can't count the number of times I've gotten frustrated with Houdini only to end up feeling stupid when I've realised it was something I did wrong.
But when it comes to nodes not functioning correctly - well, if Houdini has taught me one thing over the years, it's that if you think there's a bug in functionality, check twice, check three times, and check again… because it almost always turns out to be something stupid I did. Really, compared to every other 3D package I've used, I've never come across one where the core logic behind everything was as consistently sound and dependable as in Houdini.
It's a complicated package. It pretty much allows you to do anything imaginable - more akin to a programming language like C++ than to a conventional 3D package. But likewise it would be like saying C++ was buggy and broken. C++ *can* yield a buggy and broken program, but usually only because of flaws in the implementation you've come up with.
I know there are bugs, I don't presume Houdini to be perfect, but seriously, always triple-check your own methodology first. I can't count the number of times I've gotten frustrated with Houdini only to end up feeling stupid when I've realised it was something I did wrong.
Houdini Lounge » Real architecture project done by Houdini
- danwood82
- 48 posts
- Offline
That's absolutely stunning! Fantastic work, and interesting to know you used Houdini. How much of the design process was it used for?
Houdini Lounge » Houdini workstation
- danwood82
- 48 posts
- Offline
Yeah, for 10k, I agree that you'd probably be better off getting several machines, rather than a single super-computer. Xeons have rapidly diminishing economic returns once you get above the ~3k mark. I reckon 3x 3k workstations would likely give you a lot more aggregate power than a single 10k one. Remember that for everyday rendering/sim purposes, you don't really need much of a GPU, so you can pile the money almost entirely into CPU and RAM.
Of course, then you have to think about a file server and render distribution, etc… so it'll take a bit more homework and elbow grease getting the pipeline set up.
Of course, then you have to think about a file server and render distribution, etc… so it'll take a bit more homework and elbow grease getting the pipeline set up.
Technical Discussion » How to achieve adaptive voxel size?
- danwood82
- 48 posts
- Offline
Yeah, I think this is pretty much the holy-grail of fluid sim… I've seen a few whitepapers about adaptive-resolution fluids, but I've yet to see a production package that implements it.
Technical Discussion » Check if Points are inside a Surface
- danwood82
- 48 posts
- Offline
The group SOP will work, but it's slow for heavier bounding meshes.
If you specifically want to track whether particles are inside a fluid sim mesh, then you should store the surface VDB you use to create the mesh in the first place. Then you can just sample its value at any point to test whether that point is inside or outside of the surface.
Even if you're not testing a fluid mesh, it may ultimately be faster to use VDB from Polygons and then use that to test for inclusion.
If you specifically want to track whether particles are inside a fluid sim mesh, then you should store the surface VDB you use to create the mesh in the first place. Then you can just sample its value at any point to test whether that point is inside or outside of the surface.
Even if you're not testing a fluid mesh, it may ultimately be faster to use VDB from Polygons and then use that to test for inclusion.
Houdini Lounge » Houdini Indie - if Discontinued
- danwood82
- 48 posts
- Offline
Additionally, the hiplc files from Indie will load into both Apprentice and into full-commercial Houdini (degrading it to an Indie session in the process).
So in the event that you stop paying for a limited-commercial Indie license, or for some reason SideFX stopped offering them, you will still be able to open all your files until the end of time in a non-commercial capacity.
So in the event that you stop paying for a limited-commercial Indie license, or for some reason SideFX stopped offering them, you will still be able to open all your files until the end of time in a non-commercial capacity.
Technical Discussion » cache loading very heavy
- danwood82
- 48 posts
- Offline
igou
gtx760 4GB. I just want more memory for pyro so choose the cheaper one(have 4GB).
btw, could you give me some points that I can make sense where is the different Geforce from Quadro? the only I know is for maya quadro will faster in viewport camera orbit. and my friends say the different only appear in CAT software.
I really don't know the deep-down technical difference - I think there are sometimes cases of GeForce cards having their compute capabilities intentionally limited compared to their Quadro counterparts, but that's only potentially relevant if you're using OpenCL/CUDA capabilities, and even then, I get the feeling it only tends to apply to double-precision calculations, which I don't think Houdini or any other VFX software uses. I could be wrong, but I think double precision tends to be geared more towards scientific computing uses.
In my experience, a GTX 760 is going to fly, and I think any perceived difference in viewport performance is dubious. I've used quite a few of both, and the only times I've ever had slowdown or memory issues has been because of workstations that get packaged with a Quadro 600, which basically runs out the RAM and crashes the moment you attempt to playblast a FLIP sim :-P
The real “professional” benefits of Quadro cards come from managing a large studio, where you want to be able to roll out a driver update to 500 machines, and be reasonably confident that it won't break anything in your pipeline. It's not so much about any actual performance advantage. If you've only got one or two machines, you're not really going to save yourself much money in administration costs.
Technical Discussion » cache loading very heavy
- danwood82
- 48 posts
- Offline
igou
thank you my friend, I guess u are right. I loaded from partition0(SSD*2) but my card is a poor game card.
There isn't necessarily anything wrong with a “game card”… you'll get a lot more bang-for-buck getting a GeForce than a Quadro of the same price. Quadros can cost anything up to 6x more for essentially the same piece of hardware, but with “professional” driver support for stability. A $300 gaming card would probably work fantastically… a $300 workstation card would be about the slowest you could get.
What card do you have?
Technical Discussion » cache loading very heavy
- danwood82
- 48 posts
- Offline
Loading it is one thing, but displaying it is also a concern. If you're loading fresh geometry into the viewport every frame, and you've got a million+ polygons per frame, it can depend a on your GPU just how quickly it can be loaded onto the card.
Still, with a reasonable card, it's rather more likely that disk speed will be the bottleneck.
Still, with a reasonable card, it's rather more likely that disk speed will be the bottleneck.
Technical Discussion » Make an object float
- danwood82
- 48 posts
- Offline
If you output the “surface” field from the FLIP sim, you can sample the value of the field at any arbitrary point to get the distance to the nearest point on the fluid surface (negative distances mean the point is inside the fluid), and sample the gradient of the same field to get a normal-vector that points towards that nearest point. Multiply them together, and you get an offset that you can use to translate your object to that nearest point on the surface.
(Look for the Volume Sample and Volume Gradient nodes inside a VOPSOP)
If you then animate the point you're sampling at, you could animate the object sliding around on the surface.
This will only work nicely for relatively calm fluids… if there's lots of splashing, it'll still cause the object to jump around a lot, unless you work in some sort of multiple-point filtering or something like that.
(Look for the Volume Sample and Volume Gradient nodes inside a VOPSOP)
If you then animate the point you're sampling at, you could animate the object sliding around on the surface.
This will only work nicely for relatively calm fluids… if there's lots of splashing, it'll still cause the object to jump around a lot, unless you work in some sort of multiple-point filtering or something like that.
Houdini Lounge » Differences between Houdini Indie and Houdini Engine Indie.
- danwood82
- 48 posts
- Offline
It's $199 per year. The license will expire if you don't renew. Still a damn good deal considering the usual $4.5k + $2.5k-per-year maintenance for FX.
I don't think it's available just yet, but SESI mentioned setting up a conversion tool for apprentice-to-indie… so it may be that in the future you could drop back to apprentice if you're not using it commercially for a while, and possibly upgrade your scenes back to Indie if you wish to continue doing commercial work at a later date.
(If that happened, I presume it would take the same form as the current apprentice-to-FX conversion - ie, you send the scene files to SESI when you purchase a license, and they send back the converted files)
I don't think it's available just yet, but SESI mentioned setting up a conversion tool for apprentice-to-indie… so it may be that in the future you could drop back to apprentice if you're not using it commercially for a while, and possibly upgrade your scenes back to Indie if you wish to continue doing commercial work at a later date.
(If that happened, I presume it would take the same form as the current apprentice-to-FX conversion - ie, you send the scene files to SESI when you purchase a license, and they send back the converted files)
Houdini Lounge » Differences between Houdini Indie and Houdini Engine Indie.
- danwood82
- 48 posts
- Offline
mandrake0 covers it pretty comprehensively there.
In simplest terms… “Indie” contains “Indie Engine”.
But then “Indie Engine” does everything that “Indie” can do, just without the Houdini graphical user interface… it's command-line only, so it's good if you're only using the Engine plugin functionality, or if you're running it on a farm machine.
If you only have one computer, “Indie” is the only one you'd need.
In simplest terms… “Indie” contains “Indie Engine”.
But then “Indie Engine” does everything that “Indie” can do, just without the Houdini graphical user interface… it's command-line only, so it's good if you're only using the Engine plugin functionality, or if you're running it on a farm machine.
If you only have one computer, “Indie” is the only one you'd need.
Houdini Learning Materials » Houdini fx v/s Max
- danwood82
- 48 posts
- Offline
Maurício Cuencas
From the FX artists I've talked to, all have the same idea, that Houdini is an amazing tool, but that, for a single artist, with 3ds Max and Plugins you'll get the job done faster.
The key thing here is… Max or Maya will get the job done faster… once. Maybe even twice.
(If it's something reasonably straightforward that it's designed to do out of the box, or with a dedicated plugin… you're stuck if you want to do anything off the beaten track)
The next time you use Max or Maya on a similar task, it'll take the same amount of time… and it'll take the same amount of time the the time after that, rinse-and-repeat until you're bored witless and not an awful lot more productive :-P
The more you learn of Houdini, and the better you get to grips with your ideal workflow, the less you'll find yourself doing the same thing twice. You'll solve a problem, and next time you'll either have the solution, or you'll spend the time improving it rather than building it all over again. Every project you work on will improve and accelerate your workflow, and you'll be able to get further and further with each successive project.
You can't help but steadily build an ever more effective personal-pipeline with Houdini. That's where the real power of it lies.
Technical Discussion » Low res flip sim changes a lot if turned ti hi res
- danwood82
- 48 posts
- Offline
Efforts are always made to minimize the effect, but I'm afraid that's pretty much the nature of using any kind of fluid sim.
Effectively, the wider the separation and the fewer timesteps you use, the more “inaccurate” the result will be (the ideal real-world situation would be infinite timesteps and atomic-level resolution, while the opposite end of the resolution scale is just one giant immobile sphere representing your entire sim - obviously that will be a very poor indicator of the true sim behaviour)… so a preview-resolution sim can be pretty wide of the mark, and you should always mix in a few moderate-res tests along the way to ensure you're heading in the right direction. Going straight from super-low-preview-res to super-high-production-res will almost always leave you wondering why everything changed so much.
Depending on the speed of your fluid, one of the biggest causes can be from not increasing the substeps on the FLIP Solver node. If you double the resolution, it will be attempting to derive an accurate pressure solve while particles are moving across double the number of grid cells per substep… you'll end up getting dense mini-ripples all over your sim if the substeps are too low for a given resolution/speed combination, and they can really alter the behaviour of the sim.
Quite often, lower res sims tend to bulk out the fluid too much (as the fewer particles you have, the fatter their radius needs to be to compensate), so increasing the resolution can end up making the fluid look thinner and feel like it's covering less volume. You'll need to factor that in and sometimes emit more fluid for a high-res sim to keep it looking as full as the test.
You'll get a feel for what sort of results to expect with experience. It does require a certain amount of trial-and-error.
Effectively, the wider the separation and the fewer timesteps you use, the more “inaccurate” the result will be (the ideal real-world situation would be infinite timesteps and atomic-level resolution, while the opposite end of the resolution scale is just one giant immobile sphere representing your entire sim - obviously that will be a very poor indicator of the true sim behaviour)… so a preview-resolution sim can be pretty wide of the mark, and you should always mix in a few moderate-res tests along the way to ensure you're heading in the right direction. Going straight from super-low-preview-res to super-high-production-res will almost always leave you wondering why everything changed so much.
Depending on the speed of your fluid, one of the biggest causes can be from not increasing the substeps on the FLIP Solver node. If you double the resolution, it will be attempting to derive an accurate pressure solve while particles are moving across double the number of grid cells per substep… you'll end up getting dense mini-ripples all over your sim if the substeps are too low for a given resolution/speed combination, and they can really alter the behaviour of the sim.
Quite often, lower res sims tend to bulk out the fluid too much (as the fewer particles you have, the fatter their radius needs to be to compensate), so increasing the resolution can end up making the fluid look thinner and feel like it's covering less volume. You'll need to factor that in and sometimes emit more fluid for a high-res sim to keep it looking as full as the test.
You'll get a feel for what sort of results to expect with experience. It does require a certain amount of trial-and-error.
Houdini Lounge » Houdini for Indie Game Devs
- danwood82
- 48 posts
- Offline
Paul Geraskin
Uhh… that makes me a problem. As i will not be able to continue learning Houdini at home. And no possibility to open indie files at home either.
You can open any scene file in the free Apprentice edition just fine. The limitation is that if you save the scene file, it'll be a .hipnc Apprentice scene file, which if you then attempt to open on an Indie or Full Commercial licensed machine, will automatically downgrade them to an Apprentice session.
So, Apprentice will open any scene files for learning purposes, you just can't actually do any commercial work in it without invalidating the scene's commercial status.
Houdini Lounge » Houdini Indie rendering
- danwood82
- 48 posts
- Offline
You get a mantra token with each Engine/Batch license too… so in total, you could license a 6-machine farm, including the workstation you're using, for $900 per year. Seems like a pretty reasonable upper limit for an “indie” setup, considering each of those machines could potentially be 16-core xeons. That's a lot of potential grunt for a 1-man startup, and means licensing costs would be a tiny factor compared to the hardware costs. If you're at the point of utilising more than that plus cloud rendering could provide, I think it seems entirely reasonable to expect you to step up to the full commercial licenses :-)
Houdini Lounge » Houdini for Indie Game Devs
- danwood82
- 48 posts
- Offline
I've seen a few grumbles about the idea of a recurring annual fee, otherwise you “lose the ability to open your old work”.
There will of course always be the option of loading your old scenes into the latest version of Apprentice, even if you've stopped paying the annual subscription. You just won't be able to use those old scenes for commercial work until you renew your subscription, which seems absolutely fair and reasonable to me. It's not like you get locked out completely, you'd just get locked out commercially.
There will of course always be the option of loading your old scenes into the latest version of Apprentice, even if you've stopped paying the annual subscription. You just won't be able to use those old scenes for commercial work until you renew your subscription, which seems absolutely fair and reasonable to me. It's not like you get locked out completely, you'd just get locked out commercially.
Technical Discussion » Hologram Effect
- danwood82
- 48 posts
- Offline
That rather depends. What is a hologram effect?
Red Dwarf glued the letter H to Chris Barrie's head :-)
If you link to some reference for what look you want to try and achieve, it might yield some suggestions on how best to approach it.
Red Dwarf glued the letter H to Chris Barrie's head :-)
If you link to some reference for what look you want to try and achieve, it might yield some suggestions on how best to approach it.
Houdini Lounge » Polygon reduction / decimation?
- danwood82
- 48 posts
- Offline
I don't think there's ever going to be a remesher that gives you much for free. It's a bit like wondering how to increase the resolution of a photograph. You can apply all the filters you want, you may end up with a tidier result, but you're never going to end up with *more* information than you started with.
No remesher is going to give you UVs/parameterisation if you didn't start with any… the most it'll manage it to preserve it if you had it to begin with. If it appears to, then it's running some sort of auto-UV-unwrap or pelting after the remeshing, which you could just as easily queue up in Houdini… with all the caveats and limitations that auto-UVing/pelting carry.
No remesher is going to give you UVs/parameterisation if you didn't start with any… the most it'll manage it to preserve it if you had it to begin with. If it appears to, then it's running some sort of auto-UV-unwrap or pelting after the remeshing, which you could just as easily queue up in Houdini… with all the caveats and limitations that auto-UVing/pelting carry.
-
- Quick Links