only slightly off topic....

is there a strategy in 3D, either on the Poser level or on the higher-level apps/renderers, for making completely desaturated animations with optimized render time?

In otherwords, what if a film maker wanted to make a black and white film?

I think I could do a crude approach by making desaturated textures and making sure no colors were called in the nodes or the lights....

but would this trigger the rendering software to say "oh, grayscale...don't bother calculating color" and thus accellerate render? Or would it not save any time in the end?

I realize noir films are full of shadow and this would make the render no small project, but without color.....??

Just a wild newerbie off the wall question...

::::: Opera :::::

Message edited on: 02/15/2005 01:57

You could use grayscale materials on all the objects with or without grayscale texture maps. For example, use a medium gray diffuse color for some objects with no texture maps, and just add a bump map, and as long as you don't have any colored lights in the scene, it would render gray tonal.

If you wanted to use texture maps, you can use those already out there in color version, just convert them to grayscale in photoshop or the like. That would most likely reduce the file size of each map somewhat, which would aid in render time.

The render time difference would depend on the file size of your texture maps after being converted to grayscale, but not much else would come into effect here. Things like changing the diffuse color without any maps attached doesn't impact render time. As far as I'm aware, the renderer doesn't care if the diffuse color is black, white, or red. It's the texture maps that add on to the render.

"In otherwords, what if a film maker wanted to make a black and white film?"

The easiest way is to just shoot the scene normally, and desaturate the whole thing in post. AfterEffects, etc. Again, you could desaturate every texture of their color before hand, but that's up to you. It might not make that much of a difference in render speed, depending on the file sizes.

Message edited on: 02/15/2005 02:19

By the way, Any decent video editor application should have the ability to desaturate the entire scene very easily. It doesn't take long at all, and you probably have more control that way over the final result in the end.

I see. The most interesting idea you have there maxx is rendering with no texture...just a bump map. But besides throwing color, the texture obviously carries lots of other detail like wrinkles, etc. However, an invert of the texture (the bump map) would contain that detail too. I wonder what effect face_offs shaders would have under that setup...no texture, just a bump map, gray tones in nodes. I'm going to experiment tomorrow. ::::: Opera :::::

I would only follow that 'shoot in color, desaturate in post" if 1) I saw that going agressively grayscale saved no time and 2) suspected I would want the "colorized version" at some point. Lol.

:::: Opera :::::

Message edited on: 02/15/2005 02:35

Well, it's going to save some rendering time, but how much depends on a variety of things. With all the extra time you might spend setting up the scene for such a render would probably make up the difference. If you render in color, and desaturate in post, you'd still have the color version. The desaturated version could be saved out of your video editor as a different animation entirely. Unless you wanted something (I can't imagine what) very very specific out of doing it directly from the renderer, I just don't see it being worth the extra hassle.

VirtualDub can take a rendered colour animation and produce greyscale or monochrome video, line art, sketch effects, or even halftones, with the proper filters. And it's all free.

good points, all. Everything points to normal color render with post effects as above. The only thing that would be indicative for pure BW render would be a strategy that would save massive time, like 40%. Probably not there. ::::: Opera :::::

The effects of the threshold filter remind me a bit of Miller's *Sin City*.

i have not played much with the sketch render feature in Poser...that is...I have not done time trials. That might be another interesting avenue to explore. You can pull back the color in there, and you don't need to push the dials to radical settings where it produces "pencilized spghetti" either. ::::: Opera :::::

Which Miller? Arthur Miller? ::og::

Sin City is a BW graphic novel by Frank Miller. ;-)

Frank Miller. Same guy who revitalized Batman in The Dark Knight Returns. In the comics/graphic-novel community, he ranks up there with Alan Moore.

My favorite is "Arkham Asylum" by Grant Morrison, illustrated by the great Dave McKean. :-)

I really should write down these settings. They're creating some great line art anims.

The original unfiltered video (MPEG-1 format, 936KB) I have a terrible addiction to filters, in case it wasn't obvious ....

That line-art filter is really looking very good, LD. I have VirtualDub, I think I'll have to try it out. ;-)

If you used greyscale textures the jpeg file would be much smaller. Wouldnt that decrease render time by default?

Message edited on: 02/15/2005 05:46

yes, i think it would likos. I am going to do some time trials next weekend and will post back here. ::::: Opera :::::

There is a codec that lets you save the animation in colour or black and white but that wont save any time, the animation is still rendered in colour. Rendering in black and white (if it is possible) will probably not speed up rendering by any significant amount.

Attached Link: http://www.bauhaussoftware.com/home_LP.php

My suggestion is do the Black and white in post. If you have Photoshop, use that. You can create an action in photoshop to carry out all the filter and adjustment settings you want to get that perfect effect and then it will automatically apply it to each frame (assuming you render to stills. Some things I'd play with in photoshop Mode: Greyscale (obviously) adjust: Brightness/contrast (I'd want fairly high settings for both) Filters: blur: a very small amount of gaussian blur Artistic: Film grain again a small amount depending on our resolution If you can get a 3rd party Bloom filter ya might wanna play with that. If you don't have Photoshop already, then I would like to suggest what is my absolute favorite piece of software: Bauhaus Mirage. Think of it as Photoshop for animation, effects engine, compositor, and 2d animation tool. All with real time feed back. This software is as pricy as photoshop, but it's worth every penny (and more) for a serious poser animator. It has all the tools to do what you want, and to REALLY cut down on your render time.

Berserga, Those are good clues for Photoshop, and I am already considering increasing my skills in there and using it for my animation post processor. I quickly looked over at Bauhaus and will have to look deeper. Thanks for that lead. I am not sure how rendering in color and then post in PS or Bauhaus will cut down render time, however. In fact quite the opposite. I have the full hit for a color render, then the attention to detail in the post and the run of the stills. Meanwhile, if I find a way to get what I want in the raw initial Poser Render in grayscale, that's one (quicker, even if only somewhat) render, period. ::::: Opera :::::

When I said Mirage could save you render time (which you seem to be very concerned with) I wasn't talking about in the conversion to B&W.

One way that I often save render time in mirage is by utilizing 2d animation techniques with my 3d poser renders. Lets say I have a scene with a animated character in it and a moving camera. I do one over sized render of the scene sans character. and then render all the frames of the character and any objects he/she interacts with without background, as a series of .pngs. Keep the camera stationary.

I then import the oversized background into Mirage as a new project, and the series of character frames as a new layer. I can play at using seperate filters for foreground and background to fake things like depth of field, add effects such as fake volumetrics, lens flares, or particle effects very quickly.

Now I can save that whole oversized scene as an anim brush. and create a new project at the size of the final render (In my case it would be 853x480 for a widescreen aspect.)

Using the keyframer I can create Fake 2d camera movement by moving that anim brush. This is similar to the way 2d animatiors traditionally would use an over sized background, and Pan cels.

You can also zoom in and out and even rotate the anim brush.

In the end you have a very polished scene completed in less time you would be waiting for Poser to just finish render without any post with everything done "in camera". (especially if you used poser 5's volumetrics.)

Message edited on: 02/15/2005 11:18

One more thing ... If you render to more layers (like Foreground, mid ground, background) You have TREMENDOUS control over how your final animation will look. Remember how I said you can fake depth of field? (With Gaussian blur) Like all of Mirage's filters GB is compoletely keyframe-able so you can have fake animated change of focus. Mirage is going to be even more useful with Poser 6's shadow catcher. as it is you need to plan your shots carefully, and some shots are best done in camera.

>> *I have VirtualDub, I think I'll have to try it out. ;-)*

Here are the filter settings I used on the 21CF video, maxxmodelz.

You'll have to hunt down some of these. Filter Blender and Difference, for example, aren't included in VDub by default.

Excellent! Thanks, LD! :-)

"I am not sure how rendering in color and then post in PS or Bauhaus will cut down render time, however. In fact quite the opposite. I have the full hit for a color render, then the attention to detail in the post and the run of the stills."

Opera, if your main concern is cutting down on render time, you could still do that when using color textures. Here's a few things that will work:

1. Most texture makers save the texture at an obscenely large file size. This is to retain as much detail as possible, but for animation purposes, most of that goes to waste anyway. So one thing you can do to start with is take your original color textures and just recompress them to make the file size smaller. Here's an example:

I had a great texture by 3Dream that was originally a whopping 903KB. I recompressed it in photoshop, and got it down to 375KB without much noticable loss in quality at all, and it looks as good as the original did in my animations. I could have probably even brought it down a little more.

1. Like all your color maps, your bump maps are usually saved by the texture creators in RGB color mode. The original texture is simply desaturated, tweaked in some areas like the eyebrows and pubic hair, then saved as a large, RGB file. One thing you can do immediately to reduce the size is convert all your bump maps from RGB color to Greyscale, then recompress them again.

As an example, I took a bump map that was originally 882KB, turned it greyscale, and saved it as a JPG, and it brought it down to 222KB with almost no noticable loss in quality.

Overall, I saved a good amount of overhead by doing those simple things. Just be sure you don't overwrite the original textures... save the recompressed ones as another file name in case you aren't happy with how the results look. Even with all that overhead saved, however, the difference in rendertime wasn't extreme. Most of the render time was still consumed by P5's tendency to take absurdly long times generating shadow maps. ;-)

"Mirage is going to be even more useful with Poser 6's shadow catcher."

It would be even better if P6 has some extended G-Buffer output capability (which I'm thinking it might), so we could run a shadow pass, diffuse pass, keylight pass, specular pass, reflection pass, refraction pass, depth pass, and ambient occlusion pass and save them all as seperate layers. Then you have ultimate control in post. :-) Message edited on: 02/15/2005 13:12

Berserga That is a packed dump of information, really, it's a complete different approach to film making. I will read it over several times and absorb. Some of it I already had in mind, but much of it is new thinking. Thanks for taking time to write out the sequence like that...very helpful and provacative. You too Maxx, some of that i've heard but I am hanging on every paragraph and will put into practice. Thanks. ::::: Opera :::::

Most of the render time was still consumed by P5's tendency to take absurdly long times generating shadow maps. I've discovered that increasing the bucket size will generally help with shadow-map calculations. There's a point of diminishing returns, however, which occurs as you run out of system resources and Poser starts thrashing your hard drive. I'm not certain that using greyscale textures would boost performance. Poser still needs to upsample the textures to 24-bit (or whatever) uncompressed bitmaps before it can start rendering, right?

I love it...I come in here with a wild idea for one thing, you all present the facts and my idea turns out minimal or nothing, but I end up with a bucket full or new information, ideas, insight and direction. Amazing... ::::: Opera :::::

I'm not an animator but a couple of things occurred to me about memory and speed when it comes to rendering. First, I can't imagine how greyscale rendering would speed anything up. Poser doesn't care about the colour. Whether they're greyscale or full living colour, it still has to calculate the RGB values for each pixel. Sure, the monochrome textures take up less RAM and HD space but at rendering time there should be no gain. I then got to thinking about alternatives to textures. How about faking it with P5 materials? If you're not doing any extreme closeups, you should be able to cook up an acceptable skin by playing with shaders. Noise for bumps, or at least to get rid of the "plastic" look, then add other bits and pieces until you have something approaching skin. Disclaimer: I have absolutely no idea if this will speed up renders but I'd like to bet it would save a fair chunk of RAM.

The only problem with procedural textures is that they are calculated on the fly and therefore much slower than having a bitmap slapped on an object.

Yep, that makes sense. Shame about that. What's the gain in memory compared to a typical Hi Res texture?

it still has to calculate the RGB values for each pixel.>> yes, that was my issue at the top...is there a way to trigger the engine to wake up to the fact it is grayscale, and thus bypass all algoryths assoicated with color calc. This is such an obscure request I am fairly sure the toggle does not exist. I had very little time earlier today, but I removed the texture map and left the bump map in place, ran Real Skin Shader...nope...I don't think that is a productive path. ::::: Opera :::::

Yes, procedural textures can take longer to render than actual picture textures. Making sure that the textures are not to big for background objects and props should help optimize rendering for animations.

The memory requirement differences between procedural and bitmap textures will usually be huge. A procedural texture typically requires only several kilobytes or so, while a bitmap texture, as we all know, can be as big or small as you like. It's the usual trade-off between speed and memory, like using shadow mapped lights or ray tracing shadows.

Oops. meant to say that was directed at SamTherapy's question.

Message edited on: 02/15/2005 20:11

"Sure, the monochrome textures take up less RAM and HD space but at rendering time there should be no gain."

I would think there might be some gain in the initial loading phase if you're working with smaller file sizes, where the textures are transferred to the renderer. However, subequent to that first few seconds of load time, there should be no further gain. If the renderer translates all the maps to bitmap anyway, then there would be no gain at all. Message edited on: 02/15/2005 22:58