Jump to content

flcc

Members
  • Content count

    369
  • Donations

    0.00 CAD 
  • Joined

  • Last visited

  • Days Won

    10

flcc last won the day on October 4

flcc had the most liked content!

Community Reputation

126 Excellent

About flcc

  • Rank
    Illusionist

Personal Information

  • Name
    Fred
  • Location
    France

Recent Profile Visitors

1,762 profile views
  1. Very slow redshift render

    Strange, here it took 7s with 2x1080ti. According to octane bench 1080ti have a score of 185 and 1050ti 48 2x185/48= 7.7. Theorically your render should be 7 x 7.7 = (aproximatively) 53s it is obvious that there is something wrong, Hardware ? driver ? other ?
  2. I have worked on such things before, (not with a round angle, but that doesn't seem to be the problem). In my opinion there is a height problem in your reconstruction. It is always the same thing. Difficult to make virtual and reality coincide with large proportions. I guess you can't test live. My advice would be to make a movie where your image is still but where the "uv-project-camera" is animated and goes up in height in order to determine when, once projected in real life the perspective becomes correct. I know it's the geometry that should move, but it's easier and you can eventually transfer the difference you find to the model afterwards. In any case, this is how I proceed when there is a problem of deformation and I can't act in live. Even if you're not physically present the guy who make the test can tell you "Hey at 10s it's ok"
  3. Well Done I like this look, both curved and well shaped. It reminds me a bit of that illustrator from the 70's Chris Foss.
  4. A simpler and more efficient solution Inspired by Konstantin 's last post (one more time :)) but this time transfering distance to displacement. Embrace F3.hiplc
  5. Another try. The previous one work sometime but sometime not. Embrace F2.hiplc
  6. And some polyextrude trick ? Thanks to librarian for the ivy solver and Konstantin for the isocurve polycut trick. Embrace F.hiplc
  7. Ruler tool and measurements

    Don't forget that 44 lines means 44 edges and therefore 43 polys So 22/43 = 0.5116... Seem ok.
  8. Take a look here A bit Old, but can help.
  9. Sorry, typo error, I mean reflections and refractions. (reflexion is french word for reflection)
  10. It's not that the renders are better or faster in C4d it's that the integration is better. I find redshift in houdini a bit clunky. Just one example, in Houini/Redshift you often have to refresh manually. The result is that sometimes you don't know, if your settings have no effect or if it's redshift that doesn't refresh. Annoying and never happen in C4d. And simply put, I'm more comfortable in C4d. "So what is your workaround for reducing/eliminating noise in renders?... " As i said noise come from lighning condition, and also reflection and reflexions, not number of objects. An exterior scene with a domelight, even if your scene is very heavy, will not be a problem. In contrast, an interior scene with just some objects can be very hard to get clean. In bulk : If a reflexion generate noise you can up the sample at material level. Sometimes you have to look for things that can generate noise: for example, reflective walls when it is not necessary. Light the scene sufficiently even if it means darkening it afterwards in post production. All tips are welcome. I do a lot of post production work behind the scenes. I aslo use redgiant denoiser sometime (less now). There is many way to make optimsation in redshift, but really light is the key. And I highly recommand you to read the sample section of the redshift Doc. Very instructive. Oh, and we probably don't have the same quality criteria. As I only do animation, often in very short delays, a bit of noise here and there doesn't bother me.
  11. "the noise on the blue surface increases significantly. " I'm not an expert with optix, but denoising (noy specifically optix) can fail in some situations, when the noise is not "reconized". Probably what's happen here. "Also, when I opened your file, I got this error. Why is that? " We don't use the same version of redshift, but this should not be a problem. "I am unable to find this setting. Where is it?" "Moreover, is there any way to see a progress/percentage based timer going on while the render" Normaly at the bootom right of the render view you should see a progress bar. Maybe due to the old redshift version ? Houdini and redshift versions are pretty "linked".
  12. Funny how sometime you can bang your head for some time on an otherwise simple problem. As I say I usually make renderings with C4D. For test i had created a new Rop and Disable your, with the yellow tag. But this don't work. You MUST choose the ROP from the renderview. That's why I couldn't reduce the calculation time in Houdini. I learn something today ahah. Now the main problem was the lighting conditions. Lighting a scene with a single strong light that will generate all the light bounces is a bad idea. The result is that as there are not enough rays (the more indirect the lighting, the less rays), noise appears, which encourages to increase the overall number of samples. But in general this doesn't solve the problem unless you go very high and increase the computation time. This is also why the blue object increases the noise. The light is just above the hole and captures part of the rays. Some of them come out but in small numbers and generate noise. I invite you to have a look at the doc in the sampling section, it's really well explained. The example of keyshot, that you show, illustrates this. This a totally different lighning. Since this time there is a dome light that illuminates the scene (many rays everywhere) everything is fine. That's why I added a simple white dome in the scene. I also disabled affect specular in the ray section so as not to get the nasty white reflections. The renderings run for 9s which is reasonable (Not noise free though). On the other hand if you want noise free renders it's better to use irradiance point cloud in second GI engine. If you really want it to be super clean the Optix denoiser works pretty well. RS_RENDER QUERY F.hip
  13. I use 32 sample, all other parmater by default. For a simple scene like it could be enough, but as I said the result is VERY noisy. This is what is weird. I'll have another look tonight as it tickles me.
  14. Here is an example with the sop solver. I'm pretty sure you can also do it with MOPs Orient_fading F.hip
  15. Hello I took a look at your scene and I must say that it's a bit weird. Normally it shouldn't take this long to render. It seems that redshift doesn't like your organic shape. I've tried different kinds of optimization but nothing really works. I don't know what kind of hardware you have but at home with 2 1080ti the render takes between 45 and 47s which is huge for a scene with 3 objects and one light. The organic form is very heavy but not that heavy, and it doesn't usually affect redshift. So I exported the scene to C4D where I'm used to rendering. The render goes to 5s, which seems much more reasonable, but... is horribly noisy. This is not the usual behaviour. I don't have the time to invest any more, right now, but something is not right, not tweaking parameters problem IMHO.
×