m:studio 2.4d in use on SA Sports Hall of Fame

#281

Ahem…

That is exactly what Adaptive Antialiasing in messiah is meant to do.
The threshold you set is the difference between neighboring pixels brightness. So a threshold of 0.1 means that two pixels need to show a difference of at least 10% to get antialiased.

In XSI, I usually use a setting of 0.05 for final rendering, together with an AA between 0 and 3 (3 is used if the threshold is reached, otherwise 0)

In messiah, people often advise to use very small values like 0.015 or lower. If you think that 8bit images have a difference of 0.004 between each color, this is extremely small.
Basically such small settings go towards disabling the adaptive idea. Which may be needed for some scenes, but usually shouldn’t.
So rather try out higher thresholds with higher AA settings (3-4-5) cautiously - it may lead exactly to what you want.

Having AA settings per surface could be a nice option, but so far is implemented in no software I know of by default. For XSI there is a user hacked shader that can reduce AA for certain surfaces, but normally this is a per frame setting - and with a good adaptive algorithm, it should really work automatically.

The thing that is missing much more from messiah is good filters - this really gives the finishing touch in XSI.

But there is also an quite different solution to the problem: Sometimes it can look very good to ADD some grain afterwards. This not only can give a more filmlike look, but it also hides differences in graininess… :wink:

Cheers,

#282

Renderman does it. And as far as I know, MentalRay for Maya also allows this.

#283

:slight_smile:

This I think is my frustration. Taking the Level from 3 to 4 on threshold 0.015 increases render time from 6min to 11min. Hardly adaptive.

Here are some tests on a problematic limited region:

[ul]
[li]AA Level 3, threshold 0.015 = 0:48 (bad - lots of jaggies on the diagonal chrome)[/li][li]AA Level 4, threshold 0.05 = 1:08, Enh Hammer, Size 2.0 (fair)[/li][li]AA Level 5, threshold 0.05 = 1:37 (fair)[/li][li]AA Level 5, threshold 0.0001 = 1:46 (fair)[/li][li]AA Level 6, threshold 0.07 = 1:56 (fair)[/li][li]AA Level 7, threshold 0.2 = 2:11 (fair)[/li][/ul]best compromise…

Render times are too bad. The problem is isolated to the diagonals in the left 1/3 of image. The first render pass is so course that regardless of AA setting, there seems to be unbridgeable gaps between the bright pixels. Perhaps the solution here is not so much increased AA, but a better first render pass? Or an option for a better first pass on selected surfaces?

Here’s the first pass…

Too high a threshold also flattens out soft shadows into regions, killing the soft gradations.

So what are you / they waiting for?? :slight_smile:

#284

If Limited Region was animatable with X & Y position channels as well as X & Y size channels, and rendering with the alpha channel buffer provided a mask for the limited region, then I could manually track and render certain parts of the camera view with higher settings and then comp it over the original render.

This way I can render a AA Level 3 pass of the full frame and later an AA Level 7 pass of a limited region. Even if it not entirely the solution I need for AA, an animatable Limited Region would be really cool anyway. Many other ways I can think to use it.

#285

Yeah, animated limited region sounds cool and basically simple to implement. But I don’t think the SDK has access to it, otherwise one could build a simple tool to do it…
Are there expressions regarding the limited region?

As for AA: these extreme things are why XSI allows to define a lowest level of AA too, not just a max. So for critical things like your diagonal lines, you could set a lowest AA of 1 and a highest AA of whatever is needed.

What I don’t fully understand: why don’t you just render out the problematic objects separately and comp them later? This is easy to do and doesn’t rely on some non existing software features?

But sure your ideas are cool - don’t get me wrong here :slight_smile:

Cheers,

#286

Chrome relies on reflecting the scene around it. To render a chrome object separately I need to switch off rendering for all the other surfaces but retain their visibility in reflections, is that right? Then I could render out only the stopwatch knob at ludicrous AA settings and comp, right? So if my thinking is correct, then you’ve offered something I haven’t thought of yet. :thumbsup:

But even so, animatable Limited Region would be cool.

#287

metaNURBS.png957x576

OK, any suggestions?

The ribbon is a metaNURBS object at level 10 (see yellow viewport version at right). But when I render it renders at the equivalent of level 3. I suspect and hope that I’ve simply missed something simple :shrug:

#288

For the rendering:

You simply check “Unseen by Camera” for the objects you only want to see in reflections/refractions but not in first-generation rays from the camera.
For finer control on a per surface level, you can use TLHPro: RayType on the opacity channel of your materials (way more work but detailed control…).

For the MetaNurbs:

I fear it IS something extremely basic:
The “Tab” switch in animate mode is only metanurbing in the viewport. You shouldn’t set the values there to such high levels - it is only slowing down your interaction without much use.
I often even use the “Only draw at rendertime” switch in the MetaNurbs panel so that only on rendering, MetaNurbs are used and calculated.
For the rendersetting, you switch to Render -> Surfaces and select your object, then you toggle to the sub-tab at the bottom of the screen called “Displacement / Appearance” and set your render subdiv for that object.
This tab is extremely important anyway, since you also switch displacement on and off here and set how particles are to be rendered…

This is also the reason why people using SubdivisionSurfaces don’t need stand-in objects as much as you with your old-school approach of finely divided objects … :wink: :stuck_out_tongue:
Since subdiv objects are rather low poly normally, this works out of the box automatically. And there are many other advantages, especially with characters and bones…

Keep in mind that messiah allows finer control over subdivs than for instance XSI. The subdiv setting is used like you type it in: A setting of 4 divides each quad into 4x4 polys, a setting of 7 makes it 7x7 polys. In XSI and other software, it is a square type setting, each level of subdivision subdivides all polys once. So Level 2 is 4x4 polys already (Level 1 divides the original poly into 2x2, Level 2 divides all those resulting polys once more, so it is 4x4, Level 3 is 8x8 etc.)

So in general, you need way higher settings in messiah than in some other tools.

Cheers,

#289

Thanks again Thomas. :thumbsup:

I’ll most likely render only the stopwatch knob and then comp. I’d have to also render an alpha channel pass to mask out the ring, the ribbon and the watch body - but that’s easy enough. Custom Surface Lists would again be handy for this.

As for the subD Levels: I was right about one thing - my suspicion that I missed something simple. :blush: I’ve never really clicked that those settings are independant under the different tabs. This also would mean that a render test I did on another thread was incorrect. Well in fact those tests actually prove my mistake. :blush:

This thread is probably playing off like a comedy in places if you look at my ignorance shining through. Mr Bean of the 3D world perhaps - or am I too hard on myself? :slight_smile: I’ll console myself with the thought that somewhere out there there was someone else who didn’t know that…right? RIGHT? Anyone else?

Silence persists over the audience until the soft patter of rain on the roof becomes deafening. Finally he stops his desperate scanning for a show of hands or disgraced faces. He drops his eyes to the floor in shame as he finally comes to the painful realization that he is alone in his ignorance - he alone did not know how to set subD levels for rendering in messiah. He looks up, knowing that further humiliation is not really possible after all his prior admissions and then says with a brave smile: “Live and learn!”

#290

LOL

I wouldn’t take this too serious.
messiah isn’t a software that gives itself away too easily. Some areas are very intuitive, but others are more like an adventure game where you have to use a gem (dug out from a deserted mine, polished with sea sand in an whiskey barrel rotating on a stick for days) in bright sunlight to smelter an old candle stump and use the liquid wax as a seal for a message in a bottle to ask for help from your mom :wink:

I think you do a great job with your project and your band looks very natural as it flows over the other stuff, so congratulations to your skills instead of feeling stupid! :thumbsup:

Cheers!

#291

LOL :scream: Yup, that sort of sums up my learning process in messiah.

Thanks for the encouragement Thomas. :thumbsup: You are very kind. :slight_smile:

#292

Isn’t messiah fun? Hehe. Now this may seem a bit odd for me to suggest this. It seems the ribbon is rendering as if it were made of something very dense and if you’re going for realism, this would be a perfect opportunity to add just a touch of translucency. :slight_smile:

#293

very good idea - and depending on how close you move to the band, displacement may look very cool too… :wink:

Cheers,

#294

How would suggesting SSS on my ribbon be odd coming from you, Gary? :slight_smile:

SSS_01.png720x576

A = original image, no translucency
B = increased rendering subD level 7, displacement ON, and translucency as follows:
[ul]
[li]translucency = 0.5[/li][li]thickness = 0.01 (my ribbon’s modelled thickness is 1cm)[/li][li]light penetration = 1.0[/li][li]falloff = 20[/li][/ul]I may have overdone the translucency setting but it’s still nice for comparison. Render time only slightly increases, which is amazing. I like the increased shadow detail and the general softer more clothy feel to the ribbon.

What are your impressions / suggestions?

#295

Very good!
A bit too much SSS like you said, but much more natural looking IMO.

:thumbsup:

Gary, especially this revelation about cloth made your SSS studies very helpful for me! :thumbsup:

Cheers,

#296

I forgot to mention that translucency colour or tint was set to the colour texture map. Sensible?

Yes, Thomas, I’ll tone it down. Cool thing is that Gary’s video helped me dial in these settings in about 3 attempts. They’re not perfect, but it was painless.

Gary your input would also be appreciated since this is all your fault. :slight_smile:

#297


translucency = 0.2
AA Level = 4.0


no bump displacement = no troubling surface noise

#298

Actually, I liked the noisy surface, because that is waht gives it that cloth feeling. But if it flickers in animation it isn’t welcome of course.

#299

I’m glad my studies are helping you guys out. But, I must say, that I’m learning just as much by looking over both you and Thomas’ shoulders as well. :slight_smile:

It’s interesting that translucency softens bump/displacemet maps, essentially by filling in the once dark shadows with light. One thing you might do as Thomas suggested is to reduce the translucency. If you’re using Uniform Translucency you may want to lower the diffusion of the object to .8 (if its at 1.0) - this will darken the blown out areas a tad.

But let’s try this instead, since it’s my fault ;), change the Uniform routine to Unilimit - increase the Limit to 10 or 20. This will give somewhat of an OrenNayar effect (a more diffuse flatter effect) perfect for materials. And it will reduce the blown-out appearance common with Uniform.

I have noticed some rendering artifacts with Displacement and Translucency, but this usually only shows up when there’s a good amount of displacement. It’s also more noticeable on light colored objects. With a modest amount of displacement, I don’t think it will affect your ribbon.

#300

I’m with you exactly, except that I’m concerned about flicker. Perhaps some tweaking may find a good compromise…

Thanks for the crit :thumbsup: