Jump to content

some particles tests


slayerk

Recommended Posts

this is rendered sequence for 16 mil particles: render. 2 light sources (both shadowmap 512x512 2x2 samples). source 1 color - (1,1,0) source 2 color (0,1,1). not any fluid simulations used to derive particles. very simple setup (in pop network three nodes: source, curl noise, drag to drive particles, and two to kill out of screen particles: group by expression, kill)

further - colorcorrecting (light mixing etc...):

step01_cc.jpg

for nice paint-like looking - edge detecting:

step02_edges.jpg

Edited by slayerk
Link to comment
Share on other sites

Those are nice and you got a good watery look.

I was also inspired by this and Peter's tutorial and tried my own version.

(I'm not hijacking your thread but it's pointless to start a second identical one) So to honour you and Peter, below is my attempt.

This is 6 million particles and I don't get anywhere near 16 million. But I added a fair bit of extra attributes to the points and my machine is limited to 6Gb.

What is your hardware?

Vimeo video of animation:

.

post-4013-127622656002_thumb.jpg

post-4013-127622656959_thumb.jpg

Edited by Macha
Link to comment
Share on other sites

Hi those are nice guys!

I liked the paint look you gave it in compositing,

and the cool motion you got there without siming any fluids.

Macha - also great stuff! good touch with the colliding sphere :)

also background and lighting are impressive.

I have made a similar test recently as well, and will join you on this thread.

although that was before I have seen Peter's tut, which is really cool!

here's a 2 mill test with some fluid advection:

animation on vimeo: http://vimeo.com/12412210

Cheers,

-R

post-3649-127623341756_thumb.jpg

post-3649-127623343591_thumb.jpg

Link to comment
Share on other sites

This is 6 million particles and I don't get anywhere near 16 million. But I added a fair bit of extra attributes to the points and my machine is limited to 6Gb.

i'm kill all attributes on particles except v and Cd. I originaly planned to play around colors of particles, but couldn't get good result and then exclude Cd parameter from shader. in 16 mil i create ifd-files (it's very fast, because all geometry is delayed load and hython are not spent memory on rendertime) and render it by deadline. statistics are: peak ram usage 4 Gb. average ram usage - 2 Gb. i think in my 8 Gb PC i may increase particles count two times (also i may optimize memory by deleting not used Cd attribute from cached geometry).

theoretically i may render somthing around 32 mil particles without problems.

Edited by slayerk
Link to comment
Share on other sites

Hi,guys,very interesting and wonderful topic :rolleyes:

I wonder if this essential divide-and-conquer way for sim and layer up for rendering is the common solution in the real production?

I knew the sony imageworks employ renderman dso for interpolation on the top of really low res sim.what about the other studios ?

what's the pros and cons between the delayed loader(or krakatoa in Max)way and the DSO way or maybe the other potential ways?

Edited by ~nature~
Link to comment
Share on other sites

i'm kill all attributes on particles except v and Cd.

Do you have any opacity on it? I have and I'm thinking I don't need it. It's probably the one limiting thing now. Initially I had pscale point attributes as well because they give nice granularity but I changed them into a details attribute to save memory. In general I think the only really limiting factor is memory. Once it is in memory it renders quite fast, regardless of lights, shadow and such things. If I hit memory limit it grinds to a halt.

What I like about particles is the sharpness they give. It's such a perfect combination that I'd like to see a generic volume-particle-combination-button on the voxel fluid sims!

Link to comment
Share on other sites

I did this test 4 weeks ago, and used almost the same procedure as in the Peter Quint video. I advected particles through a low res (40 voxel) fluid sim. Then I deleted all the spare attributes on the particles and rendered them out as bgeo sequences in 25 passes. Then rendered them out using delayed load, again in 25 passes. Then in composite I turn the opacity way down for each layer and used an add blending mode on the alpha.

27.5 million particles in total.

millions of particles link

Link to comment
Share on other sites

Do you have any opacity on it? I have and I'm thinking I don't need it. It's probably the one limiting thing now. Initially I had pscale point attributes as well because they give nice granularity but I changed them into a details attribute to save memory. In general I think the only really limiting factor is memory. Once it is in memory it renders quite fast, regardless of lights, shadow and such things. If I hit memory limit it grinds to a halt.

What I like about particles is the sharpness they give. It's such a perfect combination that I'd like to see a generic volume-particle-combination-button on the voxel fluid sims!

True, my machine has 4G of RAM, and I run linux.

In order to keep the geo files as small as I can I deleted all attributes as well besides pscale, Alpha, Cd, and v of curse!

as well rendering it out as bgeo.gz which helps a lot!

Although It seems like generating Depth-Map shadows is actually slower than the render it self for me,

Ray-Trace might give slightly sharper results on the shadow but faster!

I would love seeing such a tool as well :)

Cheers,

-R

Link to comment
Share on other sites

Here Some more test!

- Well this one was is 2 mill particles, I emit them from dynamic boxes that role towards the camera and with some smoke advection at the same sim.

I attached the HIP for this one, if interested ;)

- This one is an Up-Res Smoke Advection, with turbulence and Vorticles, 3 Mill.

Cheers

-R

BoxesInk_001.hip

post-3649-12765757277_thumb.jpg

post-3649-127657573776_thumb.jpg

post-3649-127657574957_thumb.jpg

Link to comment
Share on other sites

Here Some more test!

all these examples are very intresting, but main problem of all such particle simulations is ugly grain on semi-transparent areas.

miguel salek's test (in attachment) very clear. is any ideas how to make these?

post-3849-127658390932_thumb.jpg

Link to comment
Share on other sites

I'm sure there is a better way to solve the grain problem, but what worked for me was simply more particles.

is an obvious way. how many particles needed? hundreds of millions? i'm write above - 32 mil is near to my machine maximum, and if reduce trancparency of particles, difficult to achieve the desired density.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...