Luxcore render result with blender cycle result? what i did wrong here?

General project and community related discussions and offtopic threads.
Post Reply
User avatar
richardbao
Posts: 17
Joined: Tue Mar 09, 2021 7:47 am

Luxcore render result with blender cycle result? what i did wrong here?

Post by richardbao »

the following render results were using the same scene with same HDRI lighting.
the first one was rendered with blender cycle(no AI denoise)
the second one was done with luxcore(no AI denoise) that we integrated in our 3d software.
Screenshot 2021-05-29 123217.png
Slot_5.png
the cycle one looks more clean and tone looks better.
the luxcore one looks yellowish and it feels like a filter been applied to the image.
and the image contrast looks way better on the cycle one. i was wondering if its the reason of tonemapping or color mangement?
what i did wrong on this? did i miss something while integrating the luxcore?
any suggestion would be appreciated!
CodeHD
Donor
Donor
Posts: 437
Joined: Tue Dec 11, 2018 12:38 pm
Location: Germany

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by CodeHD »

To me it looks like there are significant differences in material setup that can be addressed.

First example is the table in the front. In the cycles render, it looks smooth and like a typical table furnish. In Luxcore, it looks like bump mapping is very excessive, and it looks more like outdoor teak wood structure.

Second example is the wall behind the bed, where a lot of the contrast difference is found. Here, cycles shows a lot of texture, while in LuxCore it is flat. So maybe adjust the contrast of the shader there. A lot of light seems to be going to the left wall (which we don't see), so if the wall shader is brighter, it could lead to more scattering around the room.

Regarding tonemapping and color management, I don't really have expereicen to talk about. Maybe someone else here has practical experience with typical settings issues.
Martini
Posts: 125
Joined: Fri Nov 23, 2018 11:36 am
Location: Australia

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by Martini »

richardbao wrote: Sat May 29, 2021 4:39 am the following render results were using the same scene with same HDRI lighting.
the first one was rendered with blender cycle(no AI denoise)
the second one was done with luxcore(no AI denoise) that we integrated in our 3d software.
Can you render it with LuxCore in Blender? To rule out the possibility that the error is a setting in your 3d software.
richardbao wrote: Sat May 29, 2021 4:39 am the cycle one looks more clean and tone looks better.
the luxcore one looks yellowish and it feels like a filter been applied to the image.
To me, it looks like your LuxCore render also has way more bounces than Cycles. Can you check the light path settings are using the same number of bounces? You didn't mention whether you were using Path tracing or BiDir, if using BiDir then total bounces = light + camera.

Also Cycles render is way darker. What gamma is applied on your HDR? Check that it's 1.0 and not 2.2.
richardbao wrote: Sat May 29, 2021 4:39 am and the image contrast looks way better on the cycle one. i was wondering if its the reason of tonemapping or color mangement?
what i did wrong on this? did i miss something while integrating the luxcore?
BlendLuxCore will apply some default image pipeline to map the tone similar to how Blender does it. Maybe your software is not doing that?
richardbao wrote: Sat May 29, 2021 4:39 am any suggestion would be appreciated!
I am not an expert with engine settings - maybe this will help, maybe I am talking nonsense :lol:

Also as CodeHD has mentioned, you did not specify how you setup your materials. It looks like some have too much bump (the table), others not enough (the walls). Perhaps there are some differences there also, that may affect how much light is reflected back into the scene.

The bedside table looks the same in both though - seems like you did something right there :)


Edit: when rendering from Blender, LuxCore will log out all the scene settings (e.g. image pipeline). You could also enable the File Exporter and inspect the other settings that are being exported (e.g. materials).
AMD Ryzen Threadripper PRO 5995WX 64-Cores | 2x Gigabyte RTX 4090 Gaming OC
ASUS Pro WS WRX80E-SAGE SE WIFI | 256GB Kingston Server Premier ECC Unbuffered DDR4
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by Dade »

Cycles image has more contrast because it is nearly a direct-light-only rendering while LuxCore image has some real global illumination: the difference is really huge. I assume Cycles has one indirect bounce at best.

You should also check the color spaces/tone mapping used for LuxCore rendering, in particular, you should be careful with gamma correction: it is easy to apply double gamma correction by not correctly declaring the gamma of input images (it will "wash out" the color of the rendering). Color spaces used can can also make a really huge difference in end result. sRGB color space:

Image

Same image with ACES color space:

Image

It may be better to compare HDR Vs HDR rendering to rule out tone mapping/etc.

LuxCore render looks also a bit too sharp, I would be curious to see the pixel filtering setting. If you post a test scene blender scene may be easier to say more on the subject.
Support LuxCoreRender project with salts and bounties
User avatar
richardbao
Posts: 17
Joined: Tue Mar 09, 2021 7:47 am

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by richardbao »

Thank you guys for the suggestions.
i will take a closer look at each of them.
it will take some time for me, and i will post feedback when i am avaliable.
have a good one guys~
User avatar
richardbao
Posts: 17
Joined: Tue Mar 09, 2021 7:47 am

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by richardbao »

Dade wrote: Sat May 29, 2021 10:36 am Cycles image has more contrast because it is nearly a direct-light-only rendering while LuxCore image has some real global illumination: the difference is really huge. I assume Cycles has one indirect bounce at best.

You should also check the color spaces/tone mapping used for LuxCore rendering, in particular, you should be careful with gamma correction: it is easy to apply double gamma correction by not correctly declaring the gamma of input images (it will "wash out" the color of the rendering). Color spaces used can can also make a really huge difference in end result. sRGB color space:

Image

Same image with ACES color space:

Image

It may be better to compare HDR Vs HDR rendering to rule out tone mapping/etc.

LuxCore render looks also a bit too sharp, I would be curious to see the pixel filtering setting. If you post a test scene blender scene may be easier to say more on the subject.


Thanks for these info, a stupid question here, how can i apply ACES in luxcoreUI? or did you do that in other software?
will the OCIO support ACES and any info about this?
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by Dade »

richardbao wrote: Wed Aug 04, 2021 3:06 am how can i apply ACES in luxcoreUI? or did you do that in other software?
will the OCIO support ACES and any info about this?
The LuxCore support for OCIO is explained here: viewtopic.php?f=5&t=3446

ACES is just a set configuration files for OCIO as explained in the link above.
Support LuxCoreRender project with salts and bounties
syhrdrjrt
Posts: 6
Joined: Sat Dec 25, 2021 11:38 pm

Re: Luxcore render result with blender cycle result? what i did wrong here?

Post by syhrdrjrt »

Dade wrote: Sat May 29, 2021 10:36 am Cycles image has more contrast because it is nearly a direct-light-only rendering while LuxCore image has some real global illumination: the difference is really huge. I assume Cycles has one indirect bounce at best.
I've seen this said here several times, but I haven't seen any proof of it, it should be easy to test, right?
Post Reply