Announcement

Collapse
No announcement yet.

Exterior Animation -Day Time- Moving Objects and Camera - In the shadows...NOT GREAT

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Ah sorry, I get you now.

    Comment


    • #17
      So thanks to the help of Super Gnu, I wanted to make a easy step by step for this specific type of rendering (a moving cam, with people walking about)... This is what I have so far, I am currently rendering the results so I should be able to tell everyone how it looks in a day or so (I'll post results too, Super Gnu, If I've sitll posted the wrong stages pelase correct me lol)... The problem with me is that I have 2400 frames to render over 30 computers (during the night times) so there's not much playing around with the settings as I need to see what it looks like the next day, and I don't have long now before it needs to be given to my director) So hectic!

      1) Add all 3D moving people on to there own layer and HIDE.

      2) Calculate The LC & IM Solution - Add Irradiance Map to Primary, and Light Cache to secondary GI Engines, Set the Sub Divs to around 1500 for LC, setting the sample size to 100mm and the scale to WORLD, make sure the "Use Cam Path" is checked, Change the mode to "screen".

      In the Irradiance Map change the IM pre-set to High - Animation, Then changing it to custom to change the Min and max to -5 & -1, set the Hsphere subdivs to 150 - and interp samples to 30 - Make sure the Cam path is checked. Change the MODE to "Animation - Prepass"

      5) Render and then change "from file" in the IM section.

      6) Un-hide the 3d animated people, and use the eye dropper to select the people's materials, going into the options in the material palette and UNTICKING the "use Irradiance Map". Do this for all materials on the 3D people.

      Render using the Saved IM files (Remember to switch off the "Don't render final Image" and save a output destination.

      only thing you might consider is now, adjusting the hsph subdivs setting in the imap will control the brute force subdivs on the people. if they look grainy, crank it up.. if they look good but render really slow, drop it down. depending on scene, between 50 and 100 should do.

      thats it really.. if you want contact shadows under the feet of the people, make sure you have ao enabled in your gi settings.

      Comment


      • #18
        well thats pretty close to the method id use, but your imap settings are all out of whack. you dont need to use "animation prepass" at all.. if you do you will get a seperate imap saved for each frame, and your renders will come out really wierd.

        if you ever did a static flythrough before, its the same workflow as that... flythrough mode for lightcache (no need to fiddle much with other settings, maybe higher subdivs, no need to "use camera path")

        imap should be on "multiframe incremental" and again, no need to tick "use camera path" calculate every 10 - 30th frame depending on camera speed. if you have any doubts about method there are 1 million guides to flythrough rendering out there.



        save both the lightcache and the imap, and load BOTH when rendering. the imap will take care of all static geometry. the lightcache is needed as a basis for the brute force calculations of the animated people. without it loaded they will come out dark.

        Comment


        • #19
          I guess what I'm confused about is how can I render the LC AND IM together if I want to back burn the Calcs? or do I need to do it only on one machine? Or should I calc the LC first on one machine, then link it to the file and then Backburn the IM part to other machines?

          Comment


          • #20
            well since you only need to calculate every 20th frame (or, if its moving slowly, every 40th frame even) it shouldnt be too punishing to do one one beefy machine.. ive never had a particular issue with waiting a couple of hours for the precalc.


            however if you want to use multiple machines, why not use distributed rendering? you can even submit a job with dist. rendering enabled, to one machine through BB, and that machine will use all the ones youve seleted in dist-rendering, to help.

            the lightcache will not be distributed, it will be calculated in full on each machine, but that doesnt matter unless one machine is much slower than the others...

            the imap however will be distributed over all machines. i find you need to use slightly higher imap subdivs when distributing.. no idea why but sometimes its a bit more prone to blotchies...


            to be honest though id be surprised if you really need to do it.. you have some beefy machines there.. if you cant chew through a flythrough lightcache and imap in a few of hours on one machine, there might be an issue with your scene.


            you cannot use backburnr alone to distribute a multiframe incremental imap.... either do it on one machine, or use distributed rendering.

            Comment


            • #21
              Thank you very much for all this! I have been looking at other tutorials but sometimes it has contradictory advice, for instance in the Vray help files it says to not use a world scale in exterior animations (or something like that If I remember rightly)..

              Below are a screen shot of my settings, I'm going to hit render in a moment on just this machine, If you think there good enough? (24 threads, 24 gig ram etc..) so should eat it

              Click image for larger version

Name:	EXAMPLE1000.jpg
Views:	1
Size:	312.7 KB
ID:	844704

              Comment


              • #22
                looks fine, although as you said - screen space is better for the lightcache ( its the default setting anyway?) also - "multiframe incremental", not "incremental add to current map" although add to current map will also work, its just you dont currently have a map to incrementally add to.

                other than that, and the fact that 150 subdivs might be unnecessarily high ( did you test?) it looks good to go. 24 cores should munch through it pretty quick (depending on scene...)

                Comment


                • #23
                  Ha, thanks for spotting that! I've just been reading up on the Light Cache Sample Size, 1.0 Means it's the same size as the resolution, so 2.0 would be twice the res? I think default is 2.0 so shall I keep it to that or change to 1.0 for a nice LC?

                  Comment


                  • #24
                    leave it at default unless you have time to test.. itll work fine and is only used as the basis for the primary bounces, its never directly visible. smaller sample sizes = more blotchy..= higher subdivs needed to compensate.

                    Comment


                    • #25
                      NOTE to self, make sure I save the IM on the desktop and not onto a network drive, the saving speed for each IM file in side each other is taking longer and longer with every save lol.

                      Comment


                      • #26
                        Hey guys, what do you think about this method I found here. I tried it out and it seems to be working well...
                        http://www.mintviz.com/blog/flicker-...on-using-vray/

                        Comment


                        • #27
                          its a valid method, but ive never actually managed to get a totally clean animation out of it - particularly in difficult scenes. in a scene with tons of direct light / dense texturing, it works ok, and definitely reduces noticeable flickering, but give it an interiror or areas in shade, and it needs very high settings and a lot of testing to give a good result.

                          Comment


                          • #28
                            This actually takes me to my next point, it's almost finished the calcs, but saving to one IM file does get slower and slower, so talking generally, if I was doing a more complex animation with a lot more "Nth frames", is there a way to divide these files up a bit? and then merge them after?

                            Comment


                            • #29
                              mm.. you can do it, rendering in chunks then using imapviewer to merge them together, but ive found it tends to create very heavy imaps as it doent seem to merge them efficiently.

                              better option would be to chop up the scene and render each one as if it was a seperate camera, with a few frames overlap to hide any lighting difference there might be.


                              to be honest though imap save times have never been an issue in my jobs.

                              Comment


                              • #30
                                Thanks dude!

                                I've rendered off a bit of the animation so far so GOOD! no flickering at all and the interaction between moving people and scene looks fine so far (from afar) see attached render, note the slight noise in the shadow, might be model / material related, will investigate.

                                Click image for larger version

Name:	1.RGB_color.0263.jpg
Views:	1
Size:	131.4 KB
ID:	844708

                                Comment

                                Working...
                                X