Luminance channel equivalent?

[tangent]

In a traditional 3D rendering engine, there's a material channel called Luminance or something like that which lets you make polygons emit light. The light can be a solid color, or come from an image. I can't find an equivalent in Octane's material editor.

I ran into this trying to render a model I made of a TV. My existing model has a material for the screen area with the TV image in the Luminance channel, but when it comes into Octane, it doesn't look right because such materials don't fall into the 3 buckets Octane currently provides, diffuse, glossy, or specular.

I read that area lights don't exist in Octane yet. Since luminous materials are one way to get an area light effect in 3D programs that don't have a separate area light feature, I wouldn't be surprised if there is no way to get the TV screen effect, either. I don't actually need a GI-accurate effect here, though. I could composite the TV screen image onto the final render in Photoshop and it would look fine for my purposes. Is there a way to avoid that manual step in the current version of Octane, or do I need to just wait for proper luminance channel support?

[tungee]

Its planned for the next 2.3 release!
But it will called the emission chanel.

[radiance]

afaik you can make emitters now.

any material that reflects more light than it receives becomes an emitter/area light.
some people have done this on the forums (can't remember the posts).

basically, they stick a texture on the object's material's diffuse channel, which is a HDR .exr image, and they make the colours higher than 1.0, eg between 0 for black and 10 for white.

this will give you an area light, but, as these are only found when a ray happens to bounce on your object with that material, it's very inefficient, especially if it's a very small object in a large area.

if you use this to fake indoor lighting in a room, with small ceiling lights, you could end up rendering for 12-24 hours.
So i'd wait for 2.3 instead of making this bruteforce trick your standard workflow

Radiance

[tangent]

Good to know.

Again, my purpose isn't to use the TV screen as primary scene illumination. I just need the displayed image on the screen to show up bright and clear. If it spills some secondary light onto whatever's in front of the screen, great, but that's not actually required in my case.

Semi-related question: are there plans for features to make external compositing easier, in case I decide I don't like the way the emitter channel works in 2.3? (I worry about it being "too accurate" in my case, since I'm doing illustration type work, not photo-real work.) In C4D, I could create what it calls an object buffer for the screen's image rectangle, enable multi-pass in the rendering options, and render that object buffer out to a separate channel, creating an alpha mask for easy screen replacement in my compositing app of choice.

I did search the user manual before asking these questions. Maybe the current wish list should go into the manual, so future features show up in such searches.

[radiance]

Good to know.

Again, my purpose isn't to use the TV screen as primary scene illumination. I just need the displayed image on the screen to show up bright and clear. If it spills some secondary light onto whatever's in front of the screen, great, but that's not actually required in my case.

Semi-related question: are there plans for features to make external compositing easier, in case I decide I don't like the way the emitter channel works in 2.3? (I worry about it being "too accurate" in my case, since I'm doing illustration type work, not photo-real work.) In C4D, I could create what it calls an object buffer for the screen's image rectangle, enable multi-pass in the rendering options, and render that object buffer out to a separate channel, creating an alpha mask for easy screen replacement in my compositing app of choice.

I did search the user manual before asking these questions. Maybe the current wish list should go into the manual, so future features show up in such searches.

an alpha channel will be in the next 2.3 release, but compositing passes is something difficult to do with the limited about of video ram in current GPUs, i fear implementing that now will make it hardly used. maybe wait 6-12 months until there are cheap GPUs available with 3-4GB or more VRAM.

Radiance

[stenson]

an alpha channel will be in the next 2.3 release, but compositing passes is something difficult to do with the limited about of video ram in current GPUs, i fear implementing that now will make it hardly used. maybe wait 6-12 months until there are cheap GPUs available with 3-4GB or more VRAM.

Radiance

I agree, for now it's easy enough to output the needed passes from the host application.
But it will for sure be nice to be able to output a multi channeled full floating point open exr!

[tangent]

compositing passes is something difficult to do with the limited about of video ram in current GPUs

I'm surprised that RAM is an issue. Each pass is, at worst, just a frame buffer's worth of memory, no? And some passes, being grayscale, are 1/3 the size of a full frame buffer.

Say you're rendering 4K frames, pretty big. A full 24-bit image might be 24 MB or so, and a grayscale 8-bit channel 8 MB. A full multi-pass image set might then be:

- ambient: 24 MB
- diffuse: 24 MB
- specular: 24 MB
- shadow: 8 MB
- reflection: 24 MB
- refraction: 24 MB
- lighting (GI, etc.): 24 MB
- depth map: 8 MB
- material or object based alpha channels (requested above): 8 MB

All that takes us to just 168 MB. I sure remember when that was a lot of RAM, but not any more. Even if you have to make the full-color passes 32-bit for efficiency, you're still under 256 MB.

I thought the biggest RAM pigs in a GPU renderer like yours are meshes and textures. Actual rendered frames (or passes that get composited into frames) should be much smaller than these.

i fear implementing that now will make it hardly used.

I think you're probably right for low-end rendering projects. It's true for me, certainly.

But, isn't multi-pass rendering almost always used at the high end, things like CG feature films? I'd think the massive render farm market would be very attractive to you.

[radiance]

Hi,

Octane is a HDR render engine, it requires 3 or 4 (with alpha) floats per pixel, that's 128 bits, that's 194MB for one image in 4K full aperture cinema.
Greyscale images are HDR too, so it's 1 float per pixel, that's 48.6 MB per film.

then, ambient does'nt exist in a path tracer.
also, diffuse, specular, etc, how can this be done ? some paths are very long have have a lot of path vertices with different types.

Octane currently is not an engine that's suitable for large VFX production houses,
so why should we make it one at this time ?
It's something we're working towards, and maybe next year we'll be at a level where this is possible, when octane has what i needs to be usefull in those kinds of studios.

An alpha and z map is no issue, we can add that easily, and it's on our short term list. (alpha for 2.3)

Radiance

[Marc Tellier]

An alpha and z map is no issue, we can add that easily, and it's on our short term list. (alpha for 2.3)

That's awesome!

Marc