interior animation and some questions

[themaxxer]

hello all

I have to make my first interior animation and have some questions about quality vs. speed.

I figured out that glossies are the biggest problems. they produce most of the noise, but there's no way around glossy reflections.
the only way to get rid of the noise is AA, I have spent a whole day to get a setup that works, but now a frame renders 1.40 hours, is that passable? are there other tricks to follow?

another thing is the size to render. the anim should be 16:9 PAL, what's the exact size to set in max? is it "35mm 1.75:1 (cine)"? if yes, what's the width to set? should I render it bigger to be save if the customer wants HD later?

and the last thing. the cam is flying trough a room and turns around, so the light changes are imense. first the cam doesn't see the windows so I set the f-number to 4, but during the anim, the cam sees the windows and all gets overburned so I have to increase the f-number. but it's really hard to set this in the viewport without rendering tons of test images. the question is, do you adjust this later in post? with gammacurves or is there another way?

well a lot of questions but I am very interested what you mean. I know there are pro's out there.
thanks in advance!

best regards
themaxxer

[tammo]

I would use the clamp output options to get faster renders with glossy reflections.
For the light changes you can use reinhard color mapping, i set it to 0,3 most of the time to prevent burnouts and still get brightly lit interiors even without LWF.
16:9 PAL is 720*576 pixels with 1.46 Pixel-Aspect (non-square) 25 Frames per second. It will look distorted on your monitor but will look fine on a TV. I render 1024*576 pixels square (1,0 pixel aspect) so i get no distortion on my previews and scale it later in after effects to non-square but the render times are longer. Don't render to fields and use motion blur.

[themaxxer]

thank you very much tammo!

I know things better now. maybe I get some other replies.

best regards
themaxxer

[mik]

Hi Themaxxer

I work in advertising in france. I regularly use the parameter rendering (HDTV (video)) in 3dsmax.
Like Tammo said 1024*576 is the best settings for a 16/9 pal but if you want a 16/9 HD you must be in 1920*1080.

Sorry for my english I hope you understand;

Mickael

[themaxxer]

thanks mik

do you use any AA filter like "video" that blurs thing a bit? or do you remove noise later in after effects?

best reragrds
themaxxer

[flino2004]

Have you tried IRmap and LC with Camera path ON?
and AA 1/10 or 2/6

[pg1]

1 hour 40 mins seems high for an interior animation,(we usually have to keep it under 10 mins a frame on dual quad cores) precalulated irmap lc is probably you best option, if you got a lot of glossies then DMC anti aliasing is the way to go.
I generally use area at default as its less blurry than video, I would be careful about using motion blur in camera, it looks great but can be quiet expensive depending how you set up antialiasing.
I also tend to avoid camera moves that do big turns like your describing instead I try and divide it to numerous camera cuts. Big turns, especially in arch interiors where there is a tendency to use wide lens, can be very..... crap.
If you do decide to use multiple cameras then set the input and output animation curves to be linear to avoid the speed up slow down feeling of going from one camera to the next.
As said by other the output res is good for pal but if you can afford the render 1024 576 is a nicer option
Depending on how you plan to do post be careful changing your colour mapping away from linear.
multimatte elements are a life saver

[themaxxer]

thank you for helping me.

@pg1
I use the method you describe and divide the cam anim to several pieces then blend these in post. well the glossies are a problem. I have still noise with DMC AA 1/12 I think it's the "ward" option, or the domelight that is used with a mapped vraysky...! it produces a nicer light distribution but tends to noise. further on I use IR/LC and camera path on. maybe you can tell me more exact what AA settings you use? and what camera lens do you use? what I have to visualize is a big room, like a restaurant.

best regards
themaxxer

[joconnell]

I did a lot of broadcast arch viz stuff back in march / april with plenty of glossy materials (most floors and very light amounts on the walls) and I went with the camera path irmap / lc option. I used dmc aa - mitchell looked really nice but it was too shimmery so I went back to area. On everything I rendered a multimatte with a matte for the floor (since it was fairly reflective wood), chrome and glass. I also used a total lighting pass and a raw reflection pass. This meant if my glossy floors or chrome was too noisy, I could use the total lighting pass to remove some reflection and dial down the noise a little bit. I popped out a depth pass per camera too to do some frischluft dof in after effects which worked really nicely on the more boring areas of the house (nice for animated rack focuses).

I'd agree with pg1 that I wouldn't try to make one big camera move that goes through the entire house. If you're working for a tv program for example, you might have 40 seconds to cover the house which means that you want enough time to see each room properly (maybe 4 seconds?). If you've got a lot of rooms to cover the camera would have to move really quickly between rooms so the animation would look very hurried. Ideally use a few cuts between different rooms and use simple moves through doorways to link one room to the next so the viewer has a sense of where they are in the house.

[themaxxer]

@joconnell
wow thanks for the info.
and what's about motionblur? render into the scene or make in post, but how? and the z-depth path what zdepth min/max did you set? is save to (multichannel)exr an option?
when I work in LWF and reinhard 0.3 what do I have to set as multiplier in the totallighting element? it's so dark. and how did you used it to remove noise? what did you with the rawreflection element?

best regards
themaxxer

[themaxxer]

I have only one question now.

How should I render the anim? the only thing that's moving is the cam, so should I do it with a precalced LC in flytrough mode? if yes do I have to enable "camera path" too? and what's about IR map. use multiframe incremental or precalc it too? do I need "camera path" here?

I searched the forum, and found this:
from vlado
(*) Removed interpolation for reflections on all materials;
(*) Used animation prepass/rendering mode also with the "Use camerapath" option on;
(*) Increased the Hemispheric subdivs for the irradiance map to the default 50;
(*) Used single frame light cache with the "Use camera path" option on;
(*) Set the light cache Scale to "Screen" with 0.02 for "Sample size";
(*) Set the "Noise threshold" option in the DMC sampler to 0.005;
(*) Set the "Color threshold" for the image sampler to 0.01 ("Use DMC sampler thresh." unchecked)

but these are settings if some objects are moving.

further on I found this:

Case I - "Use light cache for glossy rays" is OFF (the default)

Phase 1 - calculating the irradiance maps
1.1. Set the light cache to "Single frame"
1.2. Optionally turn off the "Store direct light" option for the light cache (might further reduce flickering around sharp shadows e.g. by the sun)
1.3. Set the irradiance map mode to "Animation (prepass)"
1.4. Render all the animation frames to produce an irradiance map for each frame

Phase 2 - render the final animation
2.1. Turn off the light cache (set secondary engine to "None")
2.2. Set the irradiance map mode to "Animation (rendering)"
2.3. Optionally reduce the "Interpolation samples" to 10-15 and/or "Interpolation frames" to 1 to save render time
2.4. Render the final animation frames

Case II - "Use light cache for glossy rays" is ON

Phase 1 - calculating the irradiance maps
1.1. Set the light cache to "Single frame"
1.2. Optionally turn off the "Store direct light" option for the light cache
1.3. Set the irradiance map mode to "Animation (prepass)"
1.4. Render all the animation frames to produce an irradiance map for each frame

Phase 2 - render the final animation
2.1. Keep the light cache to "Single frame" mode
2.2. Set the irradiance map mode to "Animation (rendering)"
2.3. Optionally reduce the "Interpolation samples" to 10-15 and/or "Interpolation frames" to 1 to save render time
2.4. Render the final animation frames

The only difference in the two cases is that in Case I, you can turn off the light cache for the final rendering, whereas in Case II, you need to keep it to single frame.

For the final renders, I set the Secondary GI engine to "None", as I don't need the light cache then.

is this still valid?

thank you so much guys for your help!

best regards
themaxxer

[jozef_chvojka]

Hi,
I render flythroughs without moving objects like this through backburner, LWF.

Scene setup - LC for glossies OFF ( when ON it causes flickering/ noise in glossy areas )
IRRmap - MEDIUM
LC - around 1500

Light precalc - do not hide any geometry do not replace materials do not change anything ( I've had some bad experience with this in past causing long rendertime ). You adjusted your vray settings in previews already so just turn 'Don't render final image' ON and set IrrMap to 'Multiframe Incremental' mode and LightCache to 'flythrough' mode. Set paths where those maps should be saved.
Set to render every 25th frame and send to backburner. Use only 1 computer to render lights (!)...

Render image - Just turn 'Don't render final image''Don't render final image' OFF, set IRRmap to 'from file' mode and set path to this file, the same thing with LC. Set to render every image and save output files.
send to backburner use all available machines, but wait untill your lights are rendered (not neccessary, but your picture renderings MUST NOT be ahead of your light rendering - for example IRRmap and LC rendering frame 25 and animation rendering frame 26...)

I have not tested new features of Vray like that camera path precomputed lights so I don't know if there are some advantages in using them.

Regards,
Joseph

[joconnell]

@joconnell
wow thanks for the info.
and what's about motionblur? render into the scene or make in post, but how?

In most cases I didn't bother with motion blur - the cameras were very slow moving anyway so it wouldn't have made a big difference. For the budget / time I had (a week per house, 2 - 3 minutes of animation) it wasn't worth it.

and the z-depth path what zdepth min/max did you set? is save to (multichannel)exr an option?

The z depth changes depending on the camera you're using - if you've got a long hallway you've got to have a much longer zdepth value, smaller room means smaller z depth values. The handiest way I've found of doing it is actually with the regular camera - if you match a max cam to your view, you can use the environment ranges option which will give you two squares attached to the camera - if you tweak the settings so that the near is at the closest thing in your shot, and the far is at the furthest away thing in your shot, they're the values you can use for the depth channel. Admittedly a floating point image would mean you don't need to set values but that was the way I worked.

when I work in LWF and reinhard 0.3 what do I have to set as multiplier in the totallighting element? it's so dark. and how did you used it to remove noise? what did you with the rawreflection element?

I didn't find the total lighting that dark personally but it depends on the scene. If I found that the reflections were being really noisy on something like the floor for example, I'd take my matte from the multimatte and use it to put a non reflective copy of the floor over the main render - this killed off the reflections of course so I'd drop the opacity of that layer a little - it did change the look of the floor a bit but by reducing the reflections I was reducing the park of the image that was noisy so it made the render more acceptable. I didn't have to do that very often since the noise wasn't too bad. I didn't go for very soft reflections, most of the time the floor was around .85 - .95 and any very soft values were in small areas (such as chrome taps, handles) or used on less reflective objects like furniture.

Cheers!