Hi everyone,
I'm having trouble rendering several scenes that consist of a room with a lamp that is made up of 50 to 100 lightpoints.
In the first version I have to visualise a lamp that has a lot of light points encapsulated in glass.
In a second version the lights are encapsulated within a type of plastic through which the light shines, for which I use SSS.
It doesn't matter how many hours (or days!) I render both scenes, or if I use the Octane light or emissive materials on little cubes: the scene never gets rendered without thousands of fireflies.
But apart from the flies, a nice smooth render is not possible.
I know SSS takes a heavy toll on render time, but simulating the effect of a light source shing through a material with an emissive material is not very realistic.
Is this type of scene something Octane can't handle, or am I missing a different approach?
Thank you for any feedback!
How to render a scene with 50+ lights? Can Octane handle it?
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Hi Reality4,
I think i can imagine what you are talking about.
If the light is travelling through something to get to your scene then this is definitely VERY intensive.
Usually PMC will be your only option. Try increasing the caustic blur, and perhaps even using fake shadows. Also you may want to reduce the complexity of your emitters as much as possible (maybe cubes instead of spheres).
Hope that helps.
Thanks
Chris.
I think i can imagine what you are talking about.
If the light is travelling through something to get to your scene then this is definitely VERY intensive.
Usually PMC will be your only option. Try increasing the caustic blur, and perhaps even using fake shadows. Also you may want to reduce the complexity of your emitters as much as possible (maybe cubes instead of spheres).
Hope that helps.
Thanks
Chris.
Hi Fooze, thank you for the feedback.
I am planning lots of indoor animation scenes that need small artificial lights, and it seems now that they can't be rendered.
Everything renders great with a single HDR image, but night-time scenes then are very difficult.
I have nu clue about writing render engines, but wouldn't it be possible to use a different approach? It seems the current technical base falls short.
Thanks!
I am planning lots of indoor animation scenes that need small artificial lights, and it seems now that they can't be rendered.
Everything renders great with a single HDR image, but night-time scenes then are very difficult.
I have nu clue about writing render engines, but wouldn't it be possible to use a different approach? It seems the current technical base falls short.
Thanks!
Hi Reality4,
What you will need to do is tweak the scene/light sources to be as easily "find-able" by rays as possible. If your lights are behind other objects or you have complex geometry for your emitters then this will slow things down a lot. Try to make your emitting surfaces as few triangles as possible (if you can get away with a plane or cube this is better than a sphere).
If you have a material around the emitter that you can't get away with removing completely, you could try reducing it's opacity. Depending on the result this may or may not work for you.
Another thing to try is the sampling_rate, especially if you have other emitters in the scene that seem to be contributing more light than others.
Increasing the sampling rate for an emitter should help to increase it's chance to contribute to the lighting, (relative to other emitters with a lower sampling rate).
Thanks
Chris.
What you will need to do is tweak the scene/light sources to be as easily "find-able" by rays as possible. If your lights are behind other objects or you have complex geometry for your emitters then this will slow things down a lot. Try to make your emitting surfaces as few triangles as possible (if you can get away with a plane or cube this is better than a sphere).
If you have a material around the emitter that you can't get away with removing completely, you could try reducing it's opacity. Depending on the result this may or may not work for you.
Another thing to try is the sampling_rate, especially if you have other emitters in the scene that seem to be contributing more light than others.
Increasing the sampling rate for an emitter should help to increase it's chance to contribute to the lighting, (relative to other emitters with a lower sampling rate).
Thanks
Chris.
Instead of putting lights insise spheres with SSS, you might want to make the spheres emit light and control their look and light distribution with a gradient or texture. Or you could make a separate pass just to get the right look of the lamps themselves and blend the two in post. Sorry if i repeated what has been said before, haven't read the whole thread. Cheers.
Intel Core2Quad 9200 / 4 GB OCZ / Gainward 460 2GB
Can you model another basic scene showing the scenario you have problems with?
From your words I also think using a emitting durface would be better.
From your words I also think using a emitting durface would be better.
PURE3D Visualisierungen
Sys: Intel Core i9-12900K, 128GB RAM, 2x 5090 RTX, Windows 11 Pro x64, 3ds Max 2024.2
Sys: Intel Core i9-12900K, 128GB RAM, 2x 5090 RTX, Windows 11 Pro x64, 3ds Max 2024.2
Hi mbetke, jamnique and Chris,
I have attached a Max 2010 scene that is part of a much larger scene, the screenshot is a 30 minute render on a GTX590.
Imagine this scene times 50, then you would have the actual scope of my scene setup.
A reason not to use a transmission map is that it takes too much time to paint a transmission, especially on an object as this it is quite impossible.
Another is that we want to see how the lights actually penetrate the meshes, and using an unbiased renderer is of course ideal in this situation. Biased 'tricks' just don't give a realistic image. And moreover, in one scene that light output is animated or moving.
Always when I try to render indoor scenes with 10+ lights I encounter masses of fireflies.
I would like to know if it is merely render power that I need (for example the Octane cloud render system) to get a smooth firefly/grain-free render, or that Octane isn't (yet) equipped to render these kinds of scenes.
Thanks for the feedback!
I have attached a Max 2010 scene that is part of a much larger scene, the screenshot is a 30 minute render on a GTX590.
Imagine this scene times 50, then you would have the actual scope of my scene setup.
A reason not to use a transmission map is that it takes too much time to paint a transmission, especially on an object as this it is quite impossible.
Another is that we want to see how the lights actually penetrate the meshes, and using an unbiased renderer is of course ideal in this situation. Biased 'tricks' just don't give a realistic image. And moreover, in one scene that light output is animated or moving.
Always when I try to render indoor scenes with 10+ lights I encounter masses of fireflies.
I would like to know if it is merely render power that I need (for example the Octane cloud render system) to get a smooth firefly/grain-free render, or that Octane isn't (yet) equipped to render these kinds of scenes.
Thanks for the feedback!
- Attachments
-
- sss_light_test.zip
- (759.96 KiB) Downloaded 210 times
Thanks for the scene.
I made some tests and changed a bit here and there. This is what I got:
I rendered not at the pre-defined 2048px size from the file because it took a bit too long but at 1024px which was the same in a visual way.
What I did:
- Switched all lights visible = OFF
- Added invisible fill light plane to give some more ambient light. A photographer would do the same so I think it is legal and it stays unbiased.
- Changed light samples to = 32
- Changed to PMC
- Maxdepth = 8
- Caustic blur = 1
- Changed imager to Exposure = 1, Gamma = 2,2, ISO = 250, Response = Linear
On Photoshop I made a quick reduce noise pass, reduced size from 1024px to 512px (only on 2500sample image), sharpened the image slightly.
If you would put in some glare or bloom later in your "real" rendering then the noise would vanish even more I think. Depends on your scene. I guess if you would cook the image to 10000 with PMC it will be fine. With 16k even noise free.
I made some tests and changed a bit here and there. This is what I got:
I rendered not at the pre-defined 2048px size from the file because it took a bit too long but at 1024px which was the same in a visual way.
What I did:
- Switched all lights visible = OFF
- Added invisible fill light plane to give some more ambient light. A photographer would do the same so I think it is legal and it stays unbiased.
- Changed light samples to = 32
- Changed to PMC
- Maxdepth = 8
- Caustic blur = 1
- Changed imager to Exposure = 1, Gamma = 2,2, ISO = 250, Response = Linear
On Photoshop I made a quick reduce noise pass, reduced size from 1024px to 512px (only on 2500sample image), sharpened the image slightly.
If you would put in some glare or bloom later in your "real" rendering then the noise would vanish even more I think. Depends on your scene. I guess if you would cook the image to 10000 with PMC it will be fine. With 16k even noise free.
PURE3D Visualisierungen
Sys: Intel Core i9-12900K, 128GB RAM, 2x 5090 RTX, Windows 11 Pro x64, 3ds Max 2024.2
Sys: Intel Core i9-12900K, 128GB RAM, 2x 5090 RTX, Windows 11 Pro x64, 3ds Max 2024.2