[choo-chee]

Hi all, posting here since I guess it applies to both v2 and v3. is there really a way to get clean AND sharp interior renders (without using render farms)?
I'm doing all the tricks: not crossing emitters, using invisible emitters in windows, no 255,255,255 RGB .... and still after rendering for about 100 hours, I have artifacts and rough areas in my renders.
I know I can render longer but I still want to keep time tables logical and I cannot afford using render farms for every image I need... am I missing something? I'm using Maxwell from v1 to 2.7 and still have the annoying noise problem. sometimes I just render for half a week ... and still. Of course this happens mainly in interiors as exteriors are usually faster to clear. thanx for any idea ...! I really hope NL would take a sample, typical interior scene and show us what is the best way to get faster clean results.

[photomg1]

Hi Choo-Chee ,
having had a similar obsession to yours for a while . I suppose the main conclusion I came to , was just to use some noise reduction software on the bad areas .Render out out a material id/ object id pass to use for simple masks to attack those areas .The benefit of rendering out for far longer v the potential loss of quality by just using some noise reduction wasn't worth it in many situations for me. I posted an image here that was a 4 hr render plus some noise reduction as an example , near the bottom of that thread. http://www.maxwellrender.com/forum/view ... 11&t=42642.I don't think its thats bad for a single cpu non xeon machine.

The other things I would say are build it like a set when you can , and when that is not feasible just try and add some openings off camera to let more light in from the environment . Beyond that is it really that important to get rid of every last bit of noise? (its a question I've had to ask myself)


It would be nice if nextlimit was more proactive in doing some more tutorials full stop for maxwell (moving beyond just the basic how to's that they cover for every package) , I'm still waiting for mihai's coloured moving light trails from the v3 expose for example. Also think you are right it would be nice to see someone also tackle an interior in a video to show best practices etc..

Anyway I think you will drive yourself mad trying to get rid of every last bit of noise


edit : forgot to mention , I'm using an old plug in called noise ninja its far superior to what I find in adobe products for noise reduction on maxwell renders !

[feynman]

That's a good suggestion. Noise Ninja 3 is now part of Photo Ninja. And it works a treat. You should give it a try before buying more processors or resorting to render farms.

Apart from that, a concise "interior how-to" with all "best practices" would be much welcomed; also industrial and product designers have to render in interior contexts often.

[choo-chee]

thanx guys. I used to remove noise with neat image but then you have to high pass the image to get back the lines between walls and ceiling and the noise returns...
I think NL should give us a better than "just render it for gazzilion hours in a render farm" solution. I always try all sort of lighting setups and no matter what, still some noise. I 'm planning on a new test: to render a scene 3 times for the same time: first at a low 2000X1000 and try to enlarge it, then a normal 4000X2000 and then a high 6000X3000 and reduce it to the medium rez, and see what is best.

[choo-chee]

well had to render region a part of an interior and it hit me: what if you can tell maxwell that say after SL 16 is should render regions of the image, where you have more noise? like a photoshop brush or something? so instead of working on all image it will do some extra work on problematic areas?

[Nova66]

+1

I do the Render Region trick as well and would love something a little more automated.

Andrew.

[dariolanza]

Hi all,

Yes, removing noise is one of the key points for all of us.
From us we always try to work it on the internal algorithm. We also would like to be more proactive publishing tutorials and articles plenty of tips and tricks for interiors and the like, but keep in mind that, believe it or not, it is you the regular users who finally know those practical day-by-day tricks even better than us, and who we expect to share your tricks among all of us.

However, some time ago we published this tutorial that deconstructs an interior to its minimum and suggests some tricks to speed up the renders on indoor renders:

http://support.nextlimit.com/display/tu ... or+renders

We put this first stone. We would like you to take it, adapt to your particular workflow, find your particular solutions and hopefully evolve the tips described there to something better that you can share with us. Keep in mind that here at NL we do not work producing interior scenes, but focus on the technical aspects of the algorithm, so finding practical tricks or techniques is something that typically rely on the community. This happens with any software company (just think on the engineers at Adobe coding Photoshop that are not typically the guys that know the most about practical tricks for certain particular situations, which is something that external artists use to develop and share).
As I said, this is what we see in all the software communities around us (I can find many more tutorials about i.e. ZBrush created by external artists than from the Pixologic staff themselves. They produce documentation about the usage of the features, but are the everyday artists who find, develop and share the practical approaches).
Sadly our users community is for some reason more passive and they do not create blogs or forums to publish tutorials or videos with their particular techniques and tips, and always wait to be us who find those practical everyday production tricks.

We can do our best finding the best algorithms, develop new features (a custom render region customizable per object or material is under development at this moment), make some technical advises (like our suggestion of avoiding any material in Additive mode and use always Normal layers to minimize the noise), and even offer our channels to promote the contribution of anyone that wants to contribute, but keep in mind that creating a solid users community is a double-direction dialogue, where we start offering a technology and usage guidelines, and the artists share their expertise, practical approach and everyday techniques, which finally reverts in a deeper knowledge of the tool, we the developers can learn from your experience to project new features, and we all enjoy the benefits of sharing our knowledge.

Just compare the activity on other software communities (regarding the production of tutorials, blogs and spreading platforms) compared with our community to see how lonely we feel in this endeavor.
Think about how many tutorials do you consult about whatever softwares you usually use that are produced by thirty-party artists (not by the company themselves) to notice by yourselves how important their are.

However we will continue trying to share all the knowledge we can develop with our community, but we need your input for that.

I'm sure you understand.

Best regards

Dario Lanza

[seghier]

thanks Dario
creation of blogs with tutorials is very good idea . and it good if you add links of this blogs in your website
if nextlimit help users by give free upgrade of new version to who create the best tutorials

[Polyxo]

Hi Dario,

As I said, this is what we see in all the software communities around us (I can find many more tutorials about i.e. ZBrush created by external artists than from the Pixologic staff themselves. They produce documentation about the usage of the features, but are the everyday artists who find, develop and share the practical approaches).

I get your point but you really picked the most unsuitable sample. Pixologic really defeats what one can consider common sense of Software development. They don't only document each point release with >100 pdf-pages, the firm also created hundreds of free in detail demonstrations of even the most obscure feature - just check Z-Classroom. And they pump this all out alongside spectacular new tools without letting existing customers pay - for almost ten years now.
I'm not implying that NextLimit should of could do the same - I have no clue how Pixologic can do it - but they quite certainly don't fit the scheme you layed out.

[choo-chee]

thanx dario for the reply.
imho after all the most important feature that as a v2 customer will make me upgrade again, will be render times.
this is the most annoying thing in maxwell, having to wait so long for the result.
all other stuff you put in are great, however long render times are the worst part of maxwell since day 1... that is why I think it should be improved.
keep in mind that making maxwell render interiors at 25% less time means a lot, instead of 4 computers you can use 3... instead of 8 you can use 6.... instead of 16 you can use 12 .... and 50% faster ?

[dariolanza]

Hello choo-chee,

Yes, speed is the most important single feature we are constantly thinking to improve. Every single new release of Maxwell includes always a major or minor improvement in speed.
For v3, for instance, we introduced an improvement of x5 in scenes with motion blur. And continue working.
No need to say that this is the most important feature for us (speeding up is easy, the hard point is speeding up maintaining the physical accuracy).
Of course we are always working on that, you can be sure.

And Polyxo, you are right, maybe ZBrush may not be the best example, I agree. Let's think on any other company that may best represent the mayority.
But I'm sure you get the point.
We need our community to be more proactive, share their knowledge, create contents, independent blogs and platforms, etc... like we see on other communities. Mainly for their own benefit, not for ours. As I said, you are the guys that know how to work with Maxwell the most, in practical terms even better than us. You can not rely on us to produce all the contents (specially practical tips-and-tricks that one discovers on a daily basis work), the same way as the best Photoshop artists are not the ones that are writing the code.

Of course we will provide all the visibility we could offer to any independent iniciative and will continue doing all we can, but we want to reinforce that this is a double-direction dialogue, is like that for any software, except for Maxwell, whose community is more passive than others. And there are many things that only professional users can offer, while we keep on with the endeavor of coding the best engine we can.

Best

Dario Lanza