PDA

View Full Version : VPR vs. F9 - Reflection Blur differences



sampei
08-05-2014, 10:49 AM
so I was trying to set up a reflective surface, in VPR I've achieved the look that I want but then when I render the reflective surface it's very different, a lot brighter and it almost looks like what's being reflected (an image world HDR) has moved. The surface is basic, no materials or shaders. After some tests this seems to be very noticeable only with reflection blurring above zero; I do not think this is being caused by different color spaces (it's all set to linear by default). I've also tried messing with backdrop settings (reflection option is backdrop only), radiosity, polygon normals etc. so to me this seems something that has to do with the reflection blurring parameter..does anyone know more about this and possibly a method to match F9 reflections to VPR reflections? I've come up with something using incidence gradients for specularity and diffuse, but I don't think it's a very reliable way of dealing with this and the results are not exactly the same...I'm on LW10.1 by the way, haven't checked other versions.

djwaterman
08-05-2014, 12:02 PM
Your best bet is to post the scene, VPR generally mimics F9 pretty good so it's hard to know what your problem might be just from the description.

sampei
08-05-2014, 03:08 PM
here you go I've attached a zip with lwo, lws, hdr and renders..
this is from VPR
http://i.imgur.com/WTOgH1k.jpg
and from F9
http://i.imgur.com/JcM2sIs.jpg
same exact setup, I don't get it...

spherical
08-05-2014, 03:30 PM
possibly a method to match F9 reflections to VPR reflections?

Best to think of this the other way around. VPR, for all practical purposes in the foreseeable future, will only be an approximation. It's a previewer; missing a lot of capabilities of a final render engine. It's purpose is to be interactive. Doing so, it leaves out some of the more computationally intense operations. Otherwise, it would be as slow as an F9.

It has been improved a ton since it first was released. Still a long ways to go in order to completely supplant a final render. Esoteric things like Reflection Blur; especially if it's done with nodes, usually do not match; if they show up at all.

That said, I took a look. Applying 30% spec, 5% gloss and Fresnel curve gradient layers to reflection and diffuse, as opposed to having an unrealistic gradient on reflection, zero diffuse and zero spec in the original, and got results that, for all intents and purposes in a preview, pretty much match.

123515

Taking spec back to zero, which isn't real either, the two match almost exactly.

123516

Adding a Fresnel gradient to spec, they still match.

123517

At the very least, the two near-parallel planes should show similar glint. In yours, they don't, so that would be an indication that something is evidently wrong.

Thanks for posting this. I learned some stuff today.

sampei
08-05-2014, 05:17 PM
you're welcome but is it possible to replicate the VPR render that I posted? I already knew the surface was inaccurate, what I'm interested in is what changes from VPR to F9 to alter how a same surface appears so I can re-create that result and for future reference.

spherical
08-05-2014, 05:49 PM
Well, after working with the object and its textures, I wonder why you want to replicate something that is so far off by forcing a texture that doesn't respond how you expect. IOW, what are you trying to achieve? The bright glint on the lower panel and all else dark? Then I would go about it by illuminating the object differently in order to achieve that effect. Sometimes what we think doesn't look as we expect it would. I'd still approach it from a closer to real surface properties angle and work within those constraints, not put a kludge texture on the surface and hope that it'll work out. With a completely unreal texture in the mix, it makes thinking into the physics even more difficult, as no one knows what is really going on with it; it conforms to nothing.

Look at it this way. F9 is closer to reality than VPR. That is why, with an unrealistic texture, they don't match. Make the texture such that it does follow physics and they do match. We all get that it's all fake. However, there's fake and grossly fake. Working back up the food chain to replicate something that serendipitously appeared in a renderer that is farther from the truth than that which appears in the final renderer doesn't compute. Essentially, if you get an F9 to match what was seen in the original VPR, by using such a surface, the new VPR most probably won't match anymore; pretty much because the whole thing is a fluke.

Try either removing the reflection map and letting the surface reflect what it sees (I get nearly the same output as with it), then add some additional lighting to obtain the bright effect you are looking for, while altering the overall surface texture to dim down the spec/diffuse on the rest of it. Or, you could come up with a node network that limits what the surface panels see, based upon ray or incident angle. Then, I would wager that the two will match again.

djwaterman
08-05-2014, 09:45 PM
This is interesting, I loaded your scene and both VPR and F9 matched.

123519

I turned on adaptive sampling just to get a cleaner render but that's all I did. I'm using the latest version of LW.

spherical
08-05-2014, 10:12 PM
I did the same and they didn't match. Turned on AS, too, as the quality was so bad; cleaned up a lot of noise and the edges. When I loaded the model in Modeler, nothing appeared at first. Only pulling waaayy back did the object show up. Just me, but I always place models at origin, unless there are multiple models within a model that have a specifc relationship to each other and are then, obviously, not all at origin. This partial model seems to be the latter case; just threw me for a moment.

So, send the OP your set up and we're outtahere. :)

EDIT: Just loaded it up again and they DO match. When I first loaded it, I reproduced the original pair of renders. WTF is going on?! Thought it may be Draft Mode but minimal difference. Even set CS to sRGB (was at Disabled), which raised the values considerably, and they still match.

Danner
08-06-2014, 01:21 AM
Also when vpr is in "draft" mode many things are different from the final render, the shading samples, light samples, GI bounces etc. So if you want more accurate results turn draf mode off.

When I'm lighting a scene I can't use Draft mode because it will look too dark from the lack of GI bounces, and "non draft mode" is too slow... So to speed non-draft-mode up I lower Rays Per Evaluation to 100 and Secondary Bounce Rays to 30 and up the Minimum Pixel Spacing to 15. This will give me a very fast, noisy VPR but the amount of light is more accurate.

Another diference between VPR and F9 is that shadow mapped lights turn into raytraced lights in VPR, meaning there is no soft shadow and transparency works differently.

sampei
08-06-2014, 03:51 AM
thanks for the input everyone, looks like I've found another of LW10's oddities then
edit_was able to get what I wanted by adding a key to the gradient in 11.6.3 discovery where I can confirm VPR displays a correct preview :thumbsup: