Which shadow is better? Ray Trace or Depth?  For light I have two options for shadows. I was curious if one looks better than the other?

In my experience, Depth works better in Poser 4 rendering but with firefly, Ray Trace looks better.

It really depends on what you want to accomplish - and what your scene is like.

If we're not talking HUGE scenes, you could probably use adepth mapped shadow with no ill effects. Just remember th crank up the shadow map to at least 1024...

Depth mapped shadows are what's causing the notorious Glowing Nostrlis™ - you're quit of those with raytraced shadows.

The downside of raytraces shadows is that they're harder than depthmapped shadows, and if we're not talking a bright, sunny picture, they MAY look odd. IMO they're the bestest, I'm a sucker for anything raytraced, but your mileage may vary...

Oh.. on large scenes, the shadow map gets too small almost no matter what you set it to, so in that case, if you want shadows, you'll need raytraced ones.

 I'm a huge fan of the depth mapped shadows, but that's a personal thing since I try to use raytracing as infrequently as possible for speed reasons - almost exclusively for refractions, AO and some reflections.

You can use depth mapped shadows on large scenes if (viewing through the light) you can see the entire scene and the resolution is cranked up, greater than 1024 as mentioned. Depth mapped shadows can never be transparent or transmit colors the way raytraced shadows can. 

You can "blurrier" raytrace shadows if you increase the number of samples but prepared to take a hit.

If speed ain't your concern and you want no-compromise quality, Raytrace is the way to go.  And I'll take the Pepsi challenge with anyone on that.  I have experimented extensively with Depth Mapped shadows and they just don't cut it in terms of accuracy or quality, no matter what you do.

Far as rendering in P4 is concerned, there is only depth mapped, so it ain't much of a decision.

Thanks for the feedback guys.  This is good info.

 I dunno, during a fast moving motion blurred animation I think depth maps can hold their own. Feature films from Pixar / ILM didn't begin using raytrace shadows or AO until a few years ago.

Quote -  I dunno, during a fast moving motion blurred animation I think depth maps can hold their own. Feature films from Pixar / ILM didn't begin using raytrace shadows or AO until a few years ago.

Ah, but that's something different entirely.  You can get away with a lot of approximating, low-rezing and outright cheating when movement is involved.

Motion blur and the way the human brain processes movement, keyframing in some situations (the brain filling in the blanks of what literally isn't there) means things that are taken for granted in even a fairly amateurish still image are very much overkill in motion, wasting processor time and creation time.

Look at fast-moving games, like the newer FPS's.  Gorgeous graphics but the moment you stop moving and take a look around at still objects the unreality is blatantly obvious.

Quote -
Motion blur and the way the human brain processes movement, keyframing in some situations (the brain filling in the blanks of what literally isn't there) means things that are taken for granted in even a fairly amateurish still image are very much overkill in motion, wasting processor time and creation time.

I'm not sure I follow you. Please clarify.

Well for one pixar dont do humans or hyperreal stuff. They do stylised work where character and animation are given priority over making things look photoreal. That way they are able to get away with lots of cheats since the viewer expects a toonish look in the first place. Plus PRMan didnt even have raytracing till a few years back so its not as if they had the option to use raytracing.

Also to add to Keith's post, in animation renders the single most important point with lighting/shadows is consistency. ie your lighting quality should not change from frame to frame. If you can manage this, the human eye will just outright ignore any defects in the lighting itself (as the viewer would be too busy looking at the motion instead) Of course you do need a minimum level of quality, but the level is much lower then say if you were doing a still render of people or architecture, where the eye would be free to roam around and pick at various issues. And of course moblur also hides a lot of artifacts.

Pixar does not make realistic movies. Their style does not need physical accurate shadows, just shadows which work in a movie. They use ray tracing and AO as rarely as possible, since this stuff costs much more render time - and rendertime is expensive. If an effect needs rendertime (remember Sully's fur in Monster's Inc?) they just do it and spend time and money, but only if the effect is worth it. Realistic shadows and lighting would not be that much obvious when you make toon stuff, so don't expect them to spend their money there. Pixars Renderman couldn't even do ray tracing for quite a while.

If I would do a movie with poser, I would try to use the P4 renderer wherever possible, because it's faster. If it's not possible I would avoid ray tracing and AO, just to shorten render times. I would use any trick I know to make scenes simpler, use pre-rendered backgrounds wherever possible, Stuff like that. The main characters would get the render time they need, but not more. And that's exactly what pixar does - spend a lot of render time for your main effects, but save it up everywhere else.
 

But if you want to make stills and/or need the quality, go for ray traced shadows and AO. You will need it.

And what about ILM? They are certainly not known for doing hyper-real characters and they are not using the same tools as Pixar.

I'm not saying depth mapped shadows are for every situation but I think their merit is under valued. 

ILM also use PRMan as their main renderer (they only added mentalray when they had to do Star Wars episode 1) Which means that till recently they also had no choice but to use depth maps as their was no way to do raytracing fast enough. But since they did add mentalray, PRMan finally got raytracing and hardware got cheap enough, they now use it in a lot of their work.

Anyway i'm not disagreeing with you, if used correctly depth mapped shadows are fast and effective. They are particularly good in animation because they show none of the artifacts that soft raytraced shadows can, like blotchiness or noise etc. But for the majority of poser work, which consists of stills, they aren't that great.

But heck i'd be happy if more poser work used any kind of shadows at all :)

"And this is how the scene is set up ... all 5 lights are set to Point At the square in this example, with a little bit of moving around so they spread their shadow a bit more than they would if they were all at exactly the same position". Nice detectivework, pjz99. I did a smaller version in P5 to deal with the lack of blur. You must split the shadows too, no?

Use what works for you. If you like depth mapped shadows, by all means, use them. I use them as well sometimes - mainly because they render faster. When working on the materials I always use them, because they can be cached. But I generally prefer ray traced shadows, and AO.

IMHO ILM uses Photorealistic Renderman for image generation as well, so they will have to make similar decisions. There are only a few renderers which are fast and reliable enough for movie production. Nearly all Hollywood productions were done with PRenderman, Mental Ray and Lightwave, at least that were true some years ago.

 Sorry, my tone was way out of line. I think each of the three methods (though AO is not a true shadow) has merits. I agree that raytrace shadows are awesome and versatile. I don't think that they're the "end all, be all" shadow types, and should be evaluated on a shot by shot or project by project basis.

Before PRman had raytracing, those functions would be carried out by BMRT though very sophisticated message passing, but it was only used when absolutely crucial to the shot. Also I believe ILM uses mental ray not because of raytracing but due to certain PRman limitations such as rendering dielectrics very close to the camera, etc.

Quote - Nice detectivework, pjz99. I did a smaller version in P5 to deal with the lack of blur. You must split the shadows too, no?

How do you mean?

Attached Link: http://www.renderosity.com/mod/gallery/index.php?image_id=896654&member

Wow....great thread for this beginner in rendering. Thanks for all the info! I love this place. :)

No, all the shadow values are 1.0.  If you want the black area to be non-black, then 1 fill light will suffice.  Black is simply 0/0/0 color; stacking 50 shadows onto a surface won't make it any more black.  You do want to turn the brightness down though; 5 100% lights pointed at the subject will be too much light.

I had to recreate the scene since I didn't save it, but this is pretty close.  There is one fill light at 35% off to the right (spotlight again).  Also gave the ground plane a bit of displacement texture to make it more obvious what's going on.

ps: the fill light is set to cast no shadow, white color, 35% brightness.

Guess I'm missing something. How do you reduce your shadows?

With a fill light.  Do you see the single light above the camera?

pjz99 that is a really creative solution for raytraced blurred soft shadows. Thanks for sharing.

A few notes in general: for an animation you are always shooting for fast render time per frame, so exploring low settings is a frequent strategy. But with depth map shadows, beware the famous "flicker" effect at low settings and close up.

If you have any dynamic hair in the scene and want raytrace shadows, you are facing gigantic rendertime hits unless you turn off "visible to raytrace" for the hair. But if you turn off raytrace on hair your hair will not be casting shadows. What to do, what to do. Either live without shadows from hair or throw in a depth mapped light just to cast shadows from hair. Or...accept the hit.

I highly doubt that the P4 render engine is faster that Firefly....for the same effect in the scene.

I've been playing with raytrace in Carrara 6Pro, which has soft options. It works pretty well, but I found to get the really soft you have to multiply your lights and keep the settings of each low.

I am going to carry pjz idea of clustering multiple lights with only a small position difference into my Carrara trials.

::::: Opera :::::

Quote - If you have any dynamic hair in the scene and want raytrace shadows, you are facing gigantic rendertime hits unless you turn off "visible to raytrace" for the hair. But if you turn off raytrace on hair your hair will not be casting shadows. What to do, what to do. Either live without shadows from hair or throw in a depth mapped light just to cast shadows from hair. Or...accept the hit.

This is another great flaw of Poser, that you cannot set a light to only affect a selection of objects in a scene, or to ignore certain objects.  What I'd suggest for dynamic hair (and how it's done in other apps e.g. Cinema) is to have one or more specific light(s) affect hair, and other practical lights ignore it.  In Poser, all lights have a global effect for the entire scene, unless there is some trickery that can be done with materials to link a given surface with a particular light (I don't think there is).

Quote - > Quote -

Motion blur and the way the human brain processes movement, keyframing in some situations (the brain filling in the blanks of what literally isn't there) means things that are taken for granted in even a fairly amateurish still image are very much overkill in motion, wasting processor time and creation time.

I'm not sure I follow you. Please clarify.

Okay.  Imagine a 4 frame animation.  Frame 1 is a person with the arm stuck straight out, Frame 2 is the person with the sticking at a 45 degree angle up, Frame 3 has the arm straight up, Frame 4 is a copy of Frame 2.

Play those those 4 frames in a loop and the human brain sees someone waving their arm, filling in the motion required to get between those images.  You're not actually seeing the arm move at all.

Now assume that the motion being shown takes up a single second.  Use 30 frames to cover the movement.  In order to show a true representation of that movement, more detail and accuracy, you need more frames.  Say you shoot at 300 frames per second.  By definition, you've got 10 times the detail.  You have by far a more accurate depiction of the movement because you can see more detail on the exact arm position at a given time, see things like how the motion through the air affects the way the sleeve moves, the way the shadows being cast move.  Go to 3000 frames per second and it's more accurate still.

If you just want to show someone waving their arm at normal speed, it's also a complete waste of time.  All that detail is wasted because 30 frames per second is entirely adequate to show an arm moving in what looks like a lifelike and smooth way.  Adding 10 times or 100 times the detail doesn't help in the least because people simply won't see it.  Your shadows don't have to be accurate because they move too fast for people to make out details.

Instead of a person waving, imagine a person running.  The shadow on the ground while they are moving could be an indistinct blob without much definition and it won't bother people because they don't expect to make out the details anyway.  Going to all the trouble to make the shadow accurate, or reflections or highlights, is a waste of time for the same reason that shooting at 300 or 3000 FPS (and not for special effects) is a waste of time.  As long as things are consistent (lighting is coming from the same way, that sort of thing), humans are much more forgiving of lower accuracy.

On the other hand, if you want to do a render of a single instant of that movement, freezing it in place, then you do need that detail.  You can't cheat nearly as much because for a still image people will see that the shadow is wrong and won't accept an indistinct blob if there should clearly be a defined shadow.

To put it in terms of special effects, although having a film of a giant monster eating New York City is harder than rendering a single image of a giant monster eating New York City, it's easier to make the film look more realistic than a single image.  With the image you have to get the shadows all accurate, the lighting all accurate, the distance and colours accurate otherwise people will be able to pick out the inconsistencies that make it clear it's a fake.  In the film, people don't have the time to pick out those problems so the SFX artists don't have to waste their time on making it perfect.

@Keith - Ah, you articulated what I was thinking. 

richardson, could you possibly post that light set for me to look at :)**
**

2 spots and an infinite. Not much to look at...

I don't think you have enough objects in the scene, or the appropriate composition, to make shadows all that noticeable.  You could get (possibly ARE GETTING) most of your shadow effects from AO, certainly the shadow line under the arm appears to be material based AO.  I'd expect to see more shadow cast by her body on the object she's leaning against, whether from the foreground light or the very strong rim light in the background.

Quote - > Quote - If you have any dynamic hair in the scene and want raytrace shadows, you are facing gigantic rendertime hits unless you turn off "visible to raytrace" for the hair. But if you turn off raytrace on hair your hair will not be casting shadows. What to do, what to do. Either live without shadows from hair or throw in a depth mapped light just to cast shadows from hair. Or...accept the hit.

This is another great flaw of Poser, that you cannot set a light to only affect a selection of objects in a scene, or to ignore certain objects.  What I'd suggest for dynamic hair (and how it's done in other apps e.g. Cinema) is to have one or more specific light(s) affect hair, and other practical lights ignore it.  In Poser, all lights have a global effect for the entire scene, unless there is some trickery that can be done with materials to link a given surface with a particular light (I don't think there is).

this is an old thread but i had to writte what i think.
i have been readin some PIXAR papers. and pixars renderman doesnt use raytraced shadows for hair.
they created special shadows for hair and for clouds, and dust called ''deep shadows''. they are faster. with hair and dust and clouds you dont need details in the shadows. because dust and clouds alreeady blur the shadows.

i agree. it would be amazing if we could use one light only for a specific prop or hair. but it wont happen.

on the 5 lights system. i am also using this. i am praying that poser 8 will have an area light. but i am afraid that if SM will create this light the rendertime will be the same like using 5 lights. this wouldnt change anything. right?
i hope in poser 8 we will be able to make fast soft shadows.

about ILM. ILM started using AO  and raytraced shadows  in the movie Pearl Harbor. about pixar movies. pixar is sometimes using 100 lights in one enviorment. i think they dont notice the bad depth shadow maps because 100 lights hides this.

pjz99 how much blur did you use on those 5 lights. 10? 

Quote - i agree. it would be amazing if we could use one light only for a specific prop or hair. but it wont happen.

This feature is a lot more important then you realize. Generally speaking, depth mapped shadows are not a bad shadow type, it's just that they are crippled in firefly. DM shadows work best when you can do selective lighting, making each light affect only those objects that you want and ignore everything else. This way you can do shadow-only lights and get them nice and tight on each shadow caster, which gives you good results even with depth maps. But of course this is impossible in firefly so we get all these issues with DM shadows.

Also note that deep shadows are actually a modified form of DM shadows, one that takes into account transparency in the depth map and can thus be used to do nice translucent shadows on things like hair, clouds etc But this also requires that you be able to do shadow lights for the hair. Otherwise you get all the same artifacts.

i agree.
i think poser writting is to old to add all of this inside. i am reading that the new blender 2.5 will not have a lot of new features. 2.5 will be a rewritte for blender.

Quote - > Quote - i agree. it would be amazing if we could use one light only for a specific prop or hair. but it wont happen.

 Generally speaking, depth mapped shadows are not a bad shadow type, it's just that they are crippled in firefly.

so DM shadows in other softwares are better? they are bad only in poser? 

I personally prefer the look of mapped shadows , but for light through foliage , like long grass the raytraced shadows work better.  Setting a high enough map size and ensuring it covers only the rendered area is important to godd quality mapped shadows. I think its a bit subjective too.

Attached Link: http://www.cs.ucy.ac.cy/ayia-napa06/presentations/ayianapa06per.ppt

Ray traced shadows are usually much easier to set up in an realistic environment. Pixar uses DM shadows a lot, but acording to some papers they published they use hundreds of lights in a single scene. 

Take a look at the linked PPT presentation - Pixar explains why they used Ray tracing in "Cars". Some nice details inside.

Well.

I haven't been around long enough to know the limitations of Poser and render engines generally.

What is clear to me is that in real life, there is no such thing as a light that only affects certain things (e.g. hair). If one cannot get the right effect from lighting that has global effect, that tells me that something else is wrong, and I'd much rather have that fixed than see lights that can be specified to only affect particular objects in the scene.

FWIW.

This is the typical lighting setup I use lately:
http://www.renderosity.com/mod/rrfilelock/download.php?fileid=33527&key=8122

The constraints that Pixar works under have zero to do with a single user producing stills (95% of Poser use).  Feature length film productions prefer depth-mapped shadows because they render quicker than raytraced by a big margin, not because they're of higher quality.  When you have to render 30 frames per second, times 60 seconds, times 90 minutes for a typical feature film, rendering even 10% faster can mean a tremendous cost savings.  In stills, depth mapped shadows will pretty much always be of inferior quality - and while they do render faster, for your typical still frame image that's largely unimportant. 

If image quality is less valuable to you than render time, then depth mapped shadows may be what you want ^^

Quote - If one cannot get the right effect from lighting that has global effect, that tells me that something else is wrong, and I'd much rather have that fixed than see lights that can be specified to only affect particular objects in the scene.

That's the problem, Poser can't do lighting that has a real global effect (global illumination).  It behaves in a way that is fundamentally NOT realistic, and this limitation has to be worked around with many fill lights and other tricks - which create problems of their own, since you may not want a character casting shadows in three or four different directions e.g.

Please note, the lighting setup I'm showing there is not all that realistic, but I feel it's a huge improvement over the typical Poser light setup.  YMMV.

Quote - What is clear to me is that in real life, there is no such thing as a light that only affects certain things (e.g. hair). If one cannot get the right effect from lighting that has global effect, that tells me that something else is wrong, and I'd much rather have that fixed than see lights that can be specified to only affect particular objects in the scene.

If you want to compare to real life, then you only have to look at studio photography where they also do selective lighting. Of course they cant get as precise control as in CG, but they use things like gobos/barndoors, bounce cards, light diffusers and precise control of intensity to make sure that light behaves exactly how they want it to.

As CG artists, we do much the same thing, only our tools are much more 'fake' so we compensate with equally 'fake' techniques.

Quote - I have no idea if they even exist in real ray tracers, like Mental Ray or vRay.

mentalray has them, as well as deep shadows (called volumic shadows in it) Dont know about VRay.

yeah some people dont even know how many trick they use to light a set.
the actor doesnt just stand on the street and is light by the sun. the lighting is so complex like you would build it in a computer.

for example watch CSI miami. this is not realistic lighting he he :) 

but today just for the sake of it i wanted to make sharp shadows with DM shadows. i have poser pro. i did have poser 7.
this is the result that  i got. its a spot light. shadows size is 1024,blur 5 and bias 0,1.

there is no shadow around the neck. making the bias lower doesnt make it better. making the shadow map bigger also doesnt make it better.

i actually dont remember this in poser 7. so did the DM shadows got worse in poser 8? 

I tried to make a similar render. I have problems with the light in this position both with RT and DM shadows. Wierd.

I don't have time to test more. This is very strange.