This damn image is full of bugs… and after 20 hours of rendering on a P4 2.7Ghz 1Go RAM PC2700… :evil: …
The problem with VRay is that for architecturals interiors, we ALLWAYS need to use tricks, fake… like self illumated materials, standard lights, cropping bitmap, reduce bitmap blur, etc… and all of those tricks allways cose unexcped image of bugs…
Gosh !
I thought I was asymptotic to nightmare with 4 or 5 million polys, with opacities, hundreds of displace, 2 gig of bitmaps, 20 alien plugins, modeling to the centimeter with background to the kilometer, atmospherics, etc…
Then I met the guy who rendered a bathroom in 20 hours.
I love you guy !! (no offense).
You’ve got a lovely environment, even if the clothes seem to be a little “flat”. Keep up the good job.
Je ne sais pas ou tu en es de l’utilisation de Vray, mais tu pourrais peut-etre arreter les vray lights, optimiser tes settings de rendu (imap cheaps) et tweaker uniquement les gloss reflections avec des bitmaps pour economiser des iterations inutiles (ton image n’a vraiment pas besoin de 2636000 niveaux de reflection)…
-since I have used self illumated mat (yes with hdr image in both diff and self) and each time you try differents irrad map preset it gives you completly different lighting results.
- it is the same for displacement
- glass material was perfect on view A and then on view B it was dark…
- also the lighting was VERY HARD to set up to get realistic look and each light added needs a render test to check it is not overlighting the room(Lightscape was my first love… :lol: )
- antialising gives excelents results on “classics” materials while reflective ones require much more higher seetings.
The direct computation; depth = 30 is in my view way too high, thats whats causing the redness (probably) desaturate the lighting colour but bring the depth down.
allso have a look @ the ray depth of your glass & mirror Refection/refraction.
Adaptive subdivision rules! unless your doing motion blur or DOF use adaptive.
Although we all have our own preferences, there are a few noticeable differences in your settings that don’t look right to me.
Your bounce multiplier is WAY too high.
I usually have to reduce it below 1 to get a more accurate lighting and rich lighting. Because you set it that high you get some serious overblown lighting on your walls.
Your irradiance map rate settings are too low for good detailing. Use the presets in the pull down box. They are pretty good to use.
setting Hsph subdivs to 50 might be a too bit high. This parameter greatly increases the render time. I usually reduce the subdivisions, but increase the interpolation sampling. Granted, it gives a bit less detailing, but the interpolation sampling does not have the render time penalty and gives more then fair results.
You’re crazy to set the secondary bounces depth that high. :shock: Absolutely not needed !
I personally have a preference for the adaptive AA as it balances speed and accuracy on its own.
One thing i found out a while ago is, that if you use self-illuminating panels your render time will go up drastically, because they are considered to give off light. The bigger the panels, the worse it gets…
I still have a preference for using normal Max lights over Vray lights because I can tweak them better.
A rendering like you showed should not take longer then 2hrs at maximum. If not not, I usually sacrifice a bit of the detailing in raytracing and GI in favor of speed. But that my personal point of view of course..
my Hemispherical subdivisions will only be 15 - 20 at the most. then the Interpolation samples atleast 120, usually it’s over 200 to soften the samples origonaly created.
1. For the AA, use adaptive subdivision. I suppose (0, 3) will be enough.
2. Lower the secondary bounces depth, and maybe also subdivisions (maybe this is causing your render to take so long).
3. Increase the render region division values. 8x8 is way too small because each square needs some preparation, thus smaller values won’t necessarily be faster - depending on your amount of RAM, and you have enough of it.
4. Desaturate all the materials (just insisting on my opinion ).
5. Nothing to do with the render times, but try to keep the camera and it’s target at exactly the same height. This will keep the vertical lines, well … vertical, so the perception of the space you are rendering will be more accurate, not deformed, because in real life, our brain tends to see those lines vetical even if you are looking upwards or downwards. At least, try it.
And no, I don’t think that you need interpolation samples at 100 - 120, although increasing them will make the GI look softer. A good 35 or 40 are ok, IMO. The hsph subdivs are just fine the way they are, even if a 30 would do since you need to decrease the render times.
But as I already said, I like the images so keep it up :!:
Except that I see that you are using the bucket count, not size, so that should not be the source of your slowdown. That said, bucket size is important. In our tests using larger buckets speeds up rendering significantly. Going from a bucket size of 8x8 to 64x64 could cut the rendering time in half. This is because VRay renders an area around the bucket to get rid of seams, and also because each bucket needs some setup time, so the more buckets you have, the more setup time.
5. Nothing to do with the render times, but try to keep the camera and it’s target at exactly the same height. This will keep the vertical lines, well … vertical, so the perception of the space you are rendering will be more accurate, not deformed, because in real life, our brain tends to see those lines vetical even if you are looking upwards or downwards. At least, try it.
Hi Nenad,
I don’t agree with you in this.
In the daily life, we use to systematize our tools in the height between the middle of the limb and the chest. The eye tends to seek the area contains more informations. And this area is lower then the horizon. So we usually look a bit down.. hmmm… I correct myself: So *I* (and my teacher of psychology) usually look a bit down. 8)
Best regards,
Palee
p.s.: Some digital cameras can produce such an oversaturated images too (bad white balance), so i can bravely say, these pix are sooo photorealistic!
really nice images man, one thing, how did you make the lamp lights? area light? self illuminated with output? really nice result, great images i think
Thanks for all your advises but I have previously tested ALL the thinks you say and I had very bad results.
I dont know it yet but maybe I will release my scene for download for render test purpose.
The problem I think is comming from the self illumated meshes I use and also non-interpolated glossy reflection / 7 bounces and non-interplated glossy refraction.
Anyway this project was @###[@-# and the customer… too :lol:
but it is now finished
Guess what…just after finishing the whole work my main PC died (at least the motherboard)
Ill bet the customer was VERY happy It’s only anal-retentive perfectionists artists like us that will be the last one to be happy and satisfied with our work 8)
As I have said, I previously used Lightscape and even if you dont use with it photometric lights, Lightscape allways gives excellent results for interiors lighted by the exterior and/or by luminaires.
Maybe VRay will have photometric light… one day :lol:
20 hours seems WAY too excessive for such a scene. Unless you were rendering it at 4000X3000 pixels. our office finished a similar scene like that last year using the old 1.08.04 biuld. On a dual PIII 1 GHz it rendered at 800X600 in 1h45m~ . you can check it out at www.screampoint.com . the project is project 101 - Mandarin Oriental. dunno what you’re doing that off, except that you’re depth (bounces) is way too high, we usually go with 3-5.