Help Topic: OCIO and LUT

[vlado]

I have zero interest in creating a "look" in the VFB and am only interested in getting the lights to respond like they do in a camera, which does not happen in sRGB.

See my response here (which is basically that lights respond perfectly fine, thank you):
http://forums.chaosgroup.com/showthr...495#post728495

Best regards,
Vlado

[sharktacos]

Hi Vlado,

Perhaps I misspoke with saying the lights did not respond correctly. Yes the lights work just fine. What I was referring to, in my sloppy artist non-scientist way, was what is discussed in the video from Animal Logic (posted above by Recon442) https://www.youtube.com/watch?v=vKtF2S7WEv0. It shows how the colors of materials with bright lights respond differently in sRGB colorspace then they do in ACES. The basic point demonstrated in the video is that the way the colors respond to light in ACES is nicer, and closer to what a camera does. I see that as a completely separate issue from tonemapping. Afterall, the Animal Logic renders do get blownout. It's about how the colorspace affects what the material colors do when the lights get really bright.

I also want to clarify that I am not talking about doing image adjustments in the VFB. It already has plenty of controls for this IMHO. However, it is often the case that a "look" with be defined for a show in a LUT. It's helpful to be able to see the render through that look LUT in the VFB. That's a separate issue from the colorspace however. It would be good therefore to be able to read these OCIO files (defining a colorspace), and also be able to read a LUT (defining a "look" used in a film) at the same time in the VFB.

[vlado]

ACES will make the image look different, it is true, but you will still be viewing it on an sRGB monitor (unless you have one of them Rec709 monitors or whatever). Most modern digital photo cameras are calibrated for sRGB specifically (things might be different for movie cameras but I don't know details). While ACES has its uses, making your images prettier is not one of them.

Best regards,
Vlado

[vlado]

If you do want to move to ACES though, even if we did take care of all spectral colors like sun/sky/lights, you will still need to make sure all color inputs for the renderer are specified in ACES. This means that all textures and material colors must be specified in ACEScg color space. For textures, this can be done with a tool to convert them from sRGB to ACEScg, for input colors you can use VRayOCIOTex to convert them from sRGB to ACEScg. It's a lot of extra work.

Best regards,
Vlado

[LudvikKoutny]

I am probably dumb to understand that all in depth, but if I just slap one of the ACES LUTs on top of the sRGB render with all sRGB inputs, it just looks pretty. I tried it once, and now I do it in every single of my scenes.

Also, it makes phoenixFD simulations that contain fire or burning fuel look incredibly beautiful. Instead of burned out ovesaturated highlights you have this beautiful, incredibly realistic saturation and luminance falloff on the highlights of the image. I tried to match it with reinhard, but it just did not look as good no matter how I twisted it.

[vlado]

You only need a nice image and not necessarily an accurate one which is fine, but don't expect us to put that into V-Ray.

Best regards,
Vlado

[LudvikKoutny]

No need to I have LUT that works, that's all I really needed.

[sharktacos]

Recon can you elaborate on how the LUT is working for you? Are you using an ACES LUT instead of an OCIO? I thought you had said that a LUT would not work since it was 0-1...

[LudvikKoutny]

Recon can you elaborate on how the LUT is working for you? Are you using an ACES LUT instead of an OCIO? I thought you had said that a LUT would not work since it was 0-1...

LUT would not work if it was applied on sRGB, but it works as long as it's applied on Log source, which V-Ray allows for. Adanmq was kind enough to take spi1d filmic blender LUTs and convert them to .cube, which V-Ray eats. I am attaching them. I am using just one - Filmic High Contrast.cube for look development of my scenes. It just brings tones close to what cameras photograph, so translation of material properties from photos to VrayMTL by eye is a bit easier.

Filmic.zip

All credit for these goes to Adanmq.

[^Lele^]

No need to I have LUT that works, that's all I really needed.

You'll be quoted.

[sharktacos]

LUT would not work if it was applied on sRGB, but it works as long as it's applied on Log source, which V-Ray allows for. Adanmq was kind enough to take spi1d filmic blender LUTs and convert them to .cube, which V-Ray eats. I am attaching them. I am using just one - Filmic High Contrast.cube for look development of my scenes. It just brings tones close to what cameras photograph, so translation of material properties from photos to VrayMTL by eye is a bit easier.

[ATTACH]36587[/ATTACH]

All credit for these goes to Adanmq.

Thanks for the link, I'll take a look. Can you explain what you mean by applying a LUT to a log source in Vray? I'm not familiar with that. Do you perhaps mean that a log color space is used to generate the 3D LUT? I do understand that.

[LudvikKoutny]

In context of V-Ray it just means keeping the "Convert To Log Space Before Applying LUT" checkbox enabled.

[sharktacos]

In the interest of continuing this conversation, I thought I'd share these images from the whitepaper "Cinematic Color: From Your Monitor to the Big Screen." The "S-shaped tone curve" referred to in the image below (OCIO spi-animation profile) achieves an affect comparable to the purpose of the Filmic tone curve, namely "a pleasing appearance of contrast, with well balanced highlight and shadow details" in comparison to a "naive gamma 2.2 visualization (directly mapping scene-linear to display-linear), resulting in an image with low apparent contrast and poor highlight rendition."

click the image to see it full size (so you can read the text)

[sharktacos]

Also these images from here are useful in illustrating the intent of Filmic tonemapping.

(note that I say the "intent" above. I am not necessarily suggesting that Filmic be implemented, nor am I suggesting any particular solution. I am rather asking what might be a good approach to achieving the goal of tonemapping burn-out whites without saturating them, as illustrated below).

First the shot with no exposure or tone mapping applied. Note the sky is over exposed and thus white:



Next with Reinhard tonemapping. Note the saturation on the sky



Finally with Filmic tonemapping ((Haarm-Pieter Duiker’s curve, using the ALU-only version from Hable’s presentation). Note that while the bright parts of the image have been tonemapped down, they retain the non-saturated feel of brights in a photo, in comparrison with exponential or Reinhard tonemapping curves which (as the Vray help page puts it) "Saturates the colors based on their brightness."



What I would be interested in is an approach to tonemapping the bright parts of an image (burn-out) that would not make them saturated like exponential does.

[1funk]

I think it's worth mentioning that there are several different types of Reinhard tonemapping.

I come to V-Ray from modo where we had Reinhard tonemapping based on this paper: https://www.cs.utah.edu/~reinhard/cdrom/

This tonemaps based on pixel luminance which results in saturated colors (as seen in your image above). User's complained about it and we got an option to tonemap based on RGB values. This results in a "saturation to white/grey" instead. Your image loses some contrast but that can easily be added back.

As far as I can tell, V-Ray's VFB "Highlight burn" uses Reinhard tonemapping on RGB values (it matches modo Reinhard RGB tonemapper), so doesn't have the saturation problem mentioned above