PyroFx, or fumefx for a project?

[mirHadi]

hi

 

sorry If this not a very technical question but for me It is very important.

I've been using Maya for a decade and I know enough about FumeFx and .. other stuff but in Houdini I'm new and trying to keep up in a very shot period of time.

 

for a project I need a tornado and some kind of particle effects and I found Houdini very promising.

I don't know the best setting for rendering volume effect in H I'm not convinced with the result.

 

so I started quickly to see how I can manage to use PyroFx in my pipeline to get the desired effect.

not very lucky though!.

 

rendering time of Mantra for a clean, not blurry render is high compared to vray or mental ray in maya! ... scary! for a limited time and a low budget project.

you are professionals in Houdini , so please help me to choose If I can stay in H to complete my shot in Pyro or I shift back, for this time, to Maya to accomplish my task?

 

thank you in advance

 

[Guest tar]

If you are to stick with Maya, you should be using 3Delight for volumes - it simply hammers through VDBs.

 

Most likely your Mantra settings are not optimal. Can you upload something that you think is slow and perhaps people here can try to optimise it, then you can be fully informed.

[mirHadi]

marty .. thank you.

 

I will do it as soon as I'm done with some basics and intermediate settings.

but as an experience I just used pyrofx shelf tool , a single sphere and billowy smoke.

 

It is frame 41 ( the rendered image), It has enough volume samples... but.

 

- Is it possible to see the rendered time displayed in Mplayer?   so that I can compare the images.

 

 

 

 

smoke.hip

[NSugleris]

3Delight is even free too, well they used to give out 1 free render node not sure anymore.     

 

  or go here https://home.otoy.com/and click on run demo on the right side, let it load and cry when you see how fast the GPU renders volumes.   

 

Albiet gpu volume rendering is only supported in the standalone now through exported vdbs.  But when they update the houdini plugin to Octane 3 we can be more excited.   

 

In terms of simulation I went houdini and I'm never looking back, and thats even after digging into TP for Max which is incredibly powerful.    The fact Mantra is provided free and includes unlimited render nodes just sealed the deal for me        I would keep that in mind with your project whereas arnold and vray are probably per node costs, and not cheap at that.  

[anim]

...

rendering time of Mantra for a clean, not blurry render is high compared to vray or mental ray in maya! ... scary! ..

what is stopping you from rendering in Vray?

[mirHadi]

In terms of simulation I went houdini and I'm never looking back, and thats even after digging into TP for Max which is incredibly powerful.   

I never had a chance to dive into TP. as you said It is powerful.

still think Houdini is superior or what?    and I should say It is encouraging for me to dig into Houdini more and more 

 

what is stopping you from rendering in Vray?

you mean back to Maya? , I need to stick to the agreed pipeline... I'm on Houdini part and I like Houdini very much and this is a chance to jump right in the middle of hurricane..

I know It wouldn't be easy at all..   

but here I am!.

[Guest tar]

 

 

- Is it possible to see the rendered time displayed in Mplayer?   so that I can compare the images.

 

Yes, it's the 'Scheduler', available in from the Render menu in Houdini.

 

I tried to start optimising but rendering in H14 is crashing in OsX 10.11 El Capitan currently.

[NSugleris]

I never had a chance to dive into TP. as you said It is powerful.

still think Houdini is superior or what?    and I should say It is encouraging for me to dig into Houdini more and more 

 

you mean back to Maya? , I need to stick to the agreed pipeline... I'm on Houdini part and I like Houdini very much and this is a chance to jump right in the middle of hurricane..

I know It wouldn't be easy at all..   

but here I am!.

 

TP's great but the entire system being nodes from the ground up and under the hood has worked out better logically in my brain when it comes to mapping out effects now.   I'm still fighting the urge to set up systems like how I did in Maya / TP.     The last thing I was using before buying houdini was XSI's ICE which is all node based practically.  Finally I decided if I was going to do home freelance FX then it was a no brainer with Houdini's system and the infinite render nodes which turns all my computers around the house into nodes for freelance work etc. 

Edited October 10, 2015 by NSugleris

[jkunz07]

 

Here is a test I've conducted.

 

The top image is a micropoly render which I've found to be the best at producing a clean render with least noise in the quickest time possible.  The only catch is that I've changed the 'shading quality multiplier' to 2 in order to get a sharp render, otherwise volume renders with micropoly are quite blurry.

 

The middle image was rendered with the raytracing engine.  By far the fastest, rendering 1280x720 in only 11 sec!  I find raytracing gives very quick acceptable renders, but chasing the final bit of noise away can really slow it down.

 

The last image was rendered with your default mantra settings in your file.  The render engine was PBR, I tend to avoid this as raytracing is slighty quicker from my experience and produces same results.  In recent Houdini releases the volume model was introduced which causes all render engines to call PBR shader functions so I really stick with either micropoly or raytracing depending on my needs.  You had volume quality at 2 and volume shadow quality at 2!  This is extremely aggressive settings and really only needed if adding procedural noise in the shader, even in that case volume quality of 1 might be the highest you need to go.

 

The heatmaps on the right were created by enabling the render time aov and have been normalized to 0-2 seconds per bucket, so any red areas you see mean that more than 2 seconds were spent shading that bucket.

 

The render times I give could vary quite drastically depending on your CPU.

I have a 6 core i7 @ 3.8 Ghz (http://ark.intel.com/products/63697/Intel-Core-i7-3930K-Processor-12M-Cache-up-to-3_80-GHz)

My workstation at home is quite old so most people should probably expect much better render times than what I have posted.

 

Here is the file with 3 mantra nodes for each test image I've rendered.

smoke_rendertest.hip

Edited October 10, 2015 by jkunz07

[mirHadi]

thank you for your file. Jkunz07!

 

..... In recent Houdini releases the volume model was introduced which causes all render engines to call PBR shader functions so I really stick with either micropoly or raytracing depending on my needs.  

 

so It is not necessary to shift to PBR for volumes?   actually I didn't read anything about this feature.

 

the first and second render are quite fast...  very good!

a little problem is that noise are still there but mostly in contrast parts... and as you said It is too blurry. I raised the micropoly volume shadow sampling from 0.25 to 1 and noises are gone mostly.

render time jumped up ( in my machine) from 29 sec to 47 sec. clean but blurry.

 

 for raytracing, volume shadow sampling 1, It raised up from 13 sec to 22 and nearly the same quality as micropoly ... 

 

both render times are quite reasonable. but I don't know how to get rid of the blurry image.

 

 

[jkunz07]

That parameter is not volume shadow sampling, it's volume shadow quality.

It determines the step size of the ray as it marches through the volume.  So it's kind of like doing shadow calculations on a downsampled / blurred volume.  Having a small setting here like 0.2 won't introduce more noise, but it may cause flickering or other artifacts if you lower it too much.

 

With micropoly you can try raising 'shading quality multiplier' above 2 to try to get sharper volumes.  I'm not sure why micropoly is so blurry but it's unfortunate.

 

If you want to get rid of all noise with raytracing, you have to disable stochastic transparency.  Next steps would be to raise sampling on lights however your just sampling a point light so that doesn't apply here.  Increasing pixel / AA samples is probably the next thing to do.

 

You might be able to get a small speed boost out of lowering the opacity limit from 0.995 to 0.99 or lower.

[Guest tar]

with noise, you can also reduce it by using the Ray Histogram Fusion pixel filter. The default setting is too heavy IMO at  'combine -t 20.0' and makes it look like JPEG compression, but a little can go a long way. This type of noise reduction is big feature of the new renderman too..

 

Edit: another good trick is to use a noise reduction filter in post too. Like Neat Noise in Nuke. 

Edited October 10, 2015 by tar

[mirHadi]

some level of noise is quite all right to compose with live footage.. but not too much actually!.

I'll do some more tests and as I approach to my final tornado and particle system I'll put it here.

perhaps you find a better way to reduce render time and sharper quality.

 

at the meantime, if it is not too much, please provide your test renders to us beginners to learn from it..   It would be great.

thank you for your reading and effort.. and files

[Guest tar]

at the meantime, if it is not too much, please provide your test renders to us beginners to learn from it..   It would be great.

thank you for your reading and effort.. and files

 

 

jkunz07 has already uploaded some excellent work. The allegory of learning to fish, instead of being given fish is trending here on Odforce.

[Netvudu]

I agree with Marty.

I also frown when I read people talking in the name of "beginners"...unless someone elected you, it´s better to talk for yourself.

[fathom]

my experience with smoke has me using pbr with stochastic transparency, but increasing the pixel samples to 9x9+ and then turning off ray variance and only doing a single secondary ray.  then i'll pull up the stochastic samples as necessary - maybe up to 4, but often 1 will suffice.  the theory being that the high number of pixel samples will clobber any noise and with that many pixel samples you don't need any extra light sampling to smooth out your shading.

[mirHadi]

fathom thank you.

i'm doing some test with your suggestions and few others.

hope It ends up with reasonable quality and render time.

but the problem of blurriness still persists. 

[xcaseyx]

I agree with Marty.

I also frown when I read people talking in the name of "beginners"...unless someone elected you, it´s better to talk for yourself.

its a forum, people with higher post counts forget what it was like when they started posting. 

[Guest tar]

its a forum, people with higher post counts forget what it was like when they started posting. 

 

Nope - higher post count people sit here usually bewildered with the lack of effort and try to decide whether to spend their time answering posts. There are good and bad students in every field. i.e. I'm looking at the top maths students at the Art of Problem Solving forums and they kick butt. Lazy students are constantly scolded by other students. http://www.artofproblemsolving.com

Edited October 14, 2015 by tar

[carlo_c]

Just wanted to say thanks for posting this info up guys, got a pretty slow volume render of dust going at the moment and I'm pretty sure I can optimise it loads now from the info in this thread!