Alexpdennis Posted August 19, 2013 Share Posted August 19, 2013 there is a link to my new (crappy) fluid sim.... my problem is the mesh. im going after the slow mo crown splash look. I keyed the timescale attribute over about 4 frames to get the sudden slow mo. im wondering do i just do this in post/ cook/ render by decreasing the frame step to like .25 and just do it in post? BUTT, when the timescale is low (and perhaps when its not) there is this really bumpy almost poppy effect happening on the mesh. cant really figure it out. you can really see it when i put like a flat solid, non transparent shader on it. I tried changing the vdbsmoothsdf1 operation from laplachian to gaussian. that helped a little bit, but it decreased the the detail of the mesh it seemed, not in regards to reolution, but, seemingly, in the way houdini was detecting the edges. I also noticed with the gaussian setting the initial resting state of my fluid object was awesomely smooth. im wondering if there is maybe a way to blend these? or develop a technique to smooth out the mesh in general without compromising the mesh to particle interpolation that the laplachian setting yields. im still scarily bad with project optimizations and all that. but i have attached the scene file with the peetree.obj file being the file to plugin within my "fluid_from" geo node. thanks so much in advance. peetreeDish8.hipnc peetree.obj.zip Quote Link to comment Share on other sites More sharing options...
Nerox Posted August 19, 2013 Share Posted August 19, 2013 What you can do is add an 'id' attribute, this is a toggle on the FLIP solver. Write out the particles as .bgeo in SOP's. read them back in, add a Timeblend SOP and then a time warp. Using the time warp your can animate scale ramp what ever you like and it looks very smooth. I wend from a 25fps to 2000fps (like how you would shoot it in real life) slowmotion and smooth as it can be. After the timewarp you can do your meshing. Quote Link to comment Share on other sites More sharing options...
Alexpdennis Posted August 19, 2013 Author Share Posted August 19, 2013 ok so i set that up, however it looks like i have to cache out my sim with fractionalization. unless im missing something, it looks like it just holds each frame of the cache longer instead of interpolating/ tweening between each frame. is there a way to accomplish that by using the cache by 1 frame instead of a fraction? Quote Link to comment Share on other sites More sharing options...
gui Posted August 19, 2013 Share Posted August 19, 2013 with a timeblend, it should interpolate through the float frame numbers, but you can go into chops, use a chop geometry to get the animation curves, stretch them and then use a channel sop to get those curves from chops.... hope it makes sense. Quote Link to comment Share on other sites More sharing options...
Alexpdennis Posted August 19, 2013 Author Share Posted August 19, 2013 thank you so much for the reply. so im in the dark with chops.. am i calling the cached geo file in obj level, or the flip fluid object within dops? sorry im lost XD and... if im using timeblend/warp combo, am i supposed to be writing them out with substeps (on the autodop net node) to achieve subframe information for the timeblend? if not, how am i able to write that information into it and then further, will the timeblend be able to read that sh** inherently? thanks again. a Quote Link to comment Share on other sites More sharing options...
Alexpdennis Posted August 20, 2013 Author Share Posted August 20, 2013 (edited) i cannot figure this out... i have disabled the reseeding parm, tried it using a dop import instead of writing out, increasing substep, clicked the id attribute on the solver, etc... i also checked the particle count, and that doesnt change at least that i can see... still holds each frame longer instead of interpolating between them on my timeblend/ward network... im probably missing something really simple, seeing as how i'm a complete butt dumpling... any help would be greatly appreciated. you can get the same peetree dish model form my previous post thanks again yallsas crownsplash2b.hipnc Edited August 20, 2013 by Alexpdennis Quote Link to comment Share on other sites More sharing options...
mawi Posted August 20, 2013 Share Posted August 20, 2013 (edited) In your retimed_sim network the dopimport node looks for data Geometry/Visualization. The visualization doesnt contain id attribute, so change that to Geometry. Also it might be a good idea to write the particles to disk before doing the timewarp. EDIT. And as Skybar stated below. Use the timeshift SOP. I think there is a nice exampel in its helpcard. Edited August 20, 2013 by mawi Quote Link to comment Share on other sites More sharing options...
Skybar Posted August 20, 2013 Share Posted August 20, 2013 It seems to be the Timeblend that fucks it up, and you don't even need it. Just use a Timeshift with $FF/10+1, and with "Integer Frames" turned off. What this does is basically taking the current frame number ($FF - floating point frame number), dividing it by how much you want it to slow down (/10) and essentially a primitive method of what the Timeblend does (+1). What the Timeblend does is that it makes sure the first frame of your cached geo is at the first frame even after the slowdown (or whatever frame you specified). With only the Timeshift method you have to make it so that the timeshifted frame is 1 when your current frame is 1. So it doesn't always have to be +1, if you slow it down by 10 you really only need to add 0.9; so $FF/10+0.9. If you don't do this, the timeshifted simulation will start at whatever frame you divided with, so if $FF/10 - no geometry will show up until frame 10. And no, you don't really need any substeps, atleast not very many. It's all up to you though, try it without first and see if you like it. I hope that made sense. Quote Link to comment Share on other sites More sharing options...
mawi Posted August 20, 2013 Share Posted August 20, 2013 Something like this. (You need turn of reseeding in solver and render the ROP node inside the Retime network first.) However Im not sure if Timeshift or timewarp is the best for this so try both flipRetime.hipnc Quote Link to comment Share on other sites More sharing options...
Alexpdennis Posted August 20, 2013 Author Share Posted August 20, 2013 ok, first off, thank you guys soooo much for your wisdom! the timeshift with the little expression worked perfectly. i noticed that the velocity didnt get interpolated like the particle motion. which i dont really think matters for this scenario, but it'd be cool to figure out how to get that affected by the timeshift too. i then tried to rop out the sim prior to applying the timeshift and it didnt work.. but im not complaining. Quote Link to comment Share on other sites More sharing options...
Hudson Posted August 21, 2013 Share Posted August 21, 2013 (edited) And here is the velocity affected by the timeshift. Did it this morning so there are probably better ways of doing it It does compare the derivate(slope) of the timeshift with to the $FPS. That comparison is a multiplier added to the velocity attribute of the particles. Just check the slope-Chop and the VOPSOP and you get it Good luck! EDIT* I had to clean up that scene. I don't know what I was thinking when I added a motion fx and stuff flipRetime_with_vel_affect.hipnc Edited August 23, 2013 by Hudson Quote Link to comment Share on other sites More sharing options...
Alexpdennis Posted August 21, 2013 Author Share Posted August 21, 2013 man... thank you guys so much :_) Quote Link to comment Share on other sites More sharing options...
Recommended Posts
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.