Chaos Corona 13 for 3ds Max - Daily Builds Discussion

[dj_buckley]

What's going on here, I didn't post this ....

[maru]

I guess Tom accidentally edited your post instead of quoting it. Happens to the best of us! :)

[TomG]

Ahhhh apparently I hit Edit on dj_buckley's post when I meant Quote. Wondered why it seemed to be working strangely. Let me see if I can restore it, sorry about that.

Did my best to restore the original post, there's no undo, but I didn't modify the text and just took out the quoting of Alex, so I *think* I have it back to normal. So sorry!

[TomG]

Somehow it just missed us, this knowledge! Been using the default Corona HQ denoising + HQ filtering since... forever. One naturally assumes that anything native/Corona and "high quality" and which is also the default must be the best, for day-to-day use, and of course we've never had any issues with it, therefore no incentive/reason to try anything else. It's just this recent vid that caught our attention recently which caused us to change. But yes, the difference really is night-and-day. I wonder if the corona method simply needs to be improved if it's to remain the default... or just make Intel+Tent the default if better?

That's what I was repeatedly told when questioning it too, in that thread I posted you can see the other response were convinced with comments like "I rarely sharpen now with Corona defaults".  Intel + Tent is my default in my Maxstart file, i like the freedom of Tent too as you can adjust the Width(Px) to your liking, higher values softer, lower values crispier

Depends when you questioned it :) For a long time, the AI denoisers tended to blur detail and lose bump mapping, especially subtle (think wall plaster or similar), so Corona HQ was the best. That began to change in 2022 with Corona 8, "Added OIDN feature pre-denoising in the Intel AI Denoiser, which gives better preservation of detail." and Intel have made several other improvements to OIDN since then, while NVIDIA's hasn't changed at all that I know of. Over the course of the last 3 years, Intel has become by far the best denoiser, including becoming fast enough to use in IR too. From when they come visit, it also seems like they are planning to continue updating OIDN, so they aren't finished it would seem :)

[arqrenderz]

I did a post 4 years ago exposing the "corona high quality" denoiser  https://forum.corona-renderer.com/index.php?topic=32452.msg182932#msg182932
It just become my secret sauce for faster rendering, i did chat with the developers at that time and they SAY that they knew about this, corona denoiser being subpar with the intel One But meh it was never solved or looked in 4+ years.....
Feel free to render with 10% noise limit and intel denoise on 0.65
I render all our animations with these same settings, on animations i can go up to 12% noise...

[TomG]

Didn't make sense to spend time "competing" with other solutions we can include for free - even if we could compete that is, given that they a) run on GPU and b) use AI. This made trying to improve the Corona Denoiser rather unnecessary, and the dev time better spent elsewhere.

[Abdullah_Alallah]

Didn't make sense to spend time "competing" with other solutions we can include for free - even if we could compete that is, given that they a) run on GPU and b) use AI. This made trying to improve the Corona Denoiser rather unnecessary, and the dev time better spent elsewhere.

Tbh i would agree with tom here, and i think it was just a matter of time before some community members would go back to revisit to the topic of denoisers. or it would've been an industry secret that would be unfolded with time or something. so i think so far it would be just perfect to have that option back to lets say more advance users that know what needs to be done to get the quality to time and effort ratio just like the (Speed vs Accuracy balance) in the performance tab.

[Sandertype]

I did a post 4 years ago exposing the "corona high quality" denoiser  https://forum.corona-renderer.com/index.php?topic=32452.msg182932#msg182932
It just become my secret sauce for faster rendering, i did chat with the developers at that time and they SAY that they knew about this, corona denoiser being subpar with the intel One But meh it was never solved or looked in 4+ years.....
Feel free to render with 10% noise limit and intel denoise on 0.65
I render all our animations with these same settings, on animations i can go up to 12% noise...

I can't believe this is achievable since 2021! Geez.
I wonder what other secrets are hidden in this forum that would save us tons of hours and electricity.
Kind of hard to believe this wasn't used by the Corona team as default having GPU rendering winning in rendering speed. What a shame.

[TomG]

Well, just changing the denoising to Intel saves a small bit of time but not much. The time savings in the process in the video comes more from the rendering to a higher resolution and downscaling - but this does introduce artifacts at some points. Given it's "not entirely always safe", it isn't a default. As for not making the Intel Denoiser a default, this is because it is not ours and we don't control it - for example, it stopped working for some months as it was not updated for the 50 Blackwell series GPUs and simply didn't run. Again, we don't want to pick something as a default that "can break, and that is out of our control whether it does, and out of our control to fix it".

Anyway, it's historically very interesting, Corona was one of the first to introduce denoising, before other methods existed. Seems Renderman had it earlier but I don't think anything else did - according to the unreliable AI (I had to remind it after it's first answer that our denoising was in 1.4 in 2016 before it updated its answer to what is shown below):



Renderer             Denoiser Type                Release      Notes

Corona 1.4          Non-AI (Intel NLM)         2016            Early denoiser, post-process, AOV-guided
RenderMan 20    Statistical denoising       2014            One of the very first, used in film
NVIDIA OptiX      AI-based denoising        2017            First real-time AI denoiser, revolutionized workflows
Arnold                 Built-in + OptiX later       ~2017+        Integrated multiple denoising methods

[Sandertype]

Well, just changing the denoising to Intel saves a small bit of time but not much. The time savings in the process in the video comes more from the rendering to a higher resolution and downscaling - but this does introduce artifacts at some points. Given it's "not entirely always safe", it isn't a default. As for not making the Intel Denoiser a default, this is because it is not ours and we don't control it - for example, it stopped working for some months as it was not updated for the 50 Blackwell series GPUs and simply didn't run. Again, we don't want to pick something as a default that "can break, and that is out of our control whether it does, and out of our control to fix it".

It's understandable and appreciated that Chaos is careful about that. However, us, users, depend on the Corona team's R&D for workflow as well. It would be just a matter of having a warning message when using this combo.
Please, if there are other "time saving"-"quality improving" workflows out there, make them obvious for us since we're not aware of all the comments in the forum.

[TomG]

This one was news to us too :) So not something that has been kept a secret, and really it has only come to light on the creation date of that video a week or so back. Following from that video, we may add some checkbox someplace that basically does "render to 2x resolution then downscale", leaving you as a user to reduce the passes / noise target as you see fit, and apply whichever denoiser you choose though preferably Intel. We are going to revisit the denoising UI, since it has grown from one to three options and in two places (IR and final), so it could use a bit of tidying up.

TBH I am still not convinced of setting either of the AI denoisers because a new user could load up Corona and press render... . and it doesn't work. I get that established users will go "Oh, looks like Intel hasn't updated to handle the new 60 series card that I am using, I guess I will wait for an OIDN update, meantime I'll swap to another denoiser", but for a new user, the whole "It just works out of the box" statement would seem like a lie and they'd be left very confused.

There's also the issue of "reset to defaults" being something that should fix ANY problem, and same thing again, if NVIDIA or Intel break their denoisers somehow at any time, then reset to defaults will just be resetting things to something that doesn't work.

You can of course set your own default scene that loads on startup of the DCC, and in that store the settings you want to use every time, all without us having to change what the Corona defaults are. And this is my preferred solution, as then the defaults are the things guaranteed to work, but you are free to create your own standard settings via a default startup scene - best of both worlds!

[maru]

My argument against setting any denoiser as a default would be that most likely you would get flickering when rendering sequences.

Corona defaults are supposed to be safe and with as little bias as possible while keeping reasonable render times.

Also, consider these posts:

There are limitations though which aren't obvious:
- Complicated shader trees tend to resolve detail (bump, procedural details in noise maps etc) at a later stage in the render progress. With only a few passes the result will be less surface details.
- Some of Corona's sampling-based maps like AO and Curvature will also refine at later stages. I got some heavy differences between the two images where AO was much less pronounced and detailed in the downsampled image (with less passes). If you're using those kind of maps, it's best to still allow more samples.
- Refraction details also tend to take a bit longer, so you trade sharper details (implying more detail) for less actual detail and accuracy.

(...)We're finding significant detail boost at lower passes using this method, and that's without actually rendering at 2x scale and downscaling back down.(...).

So maybe that video isn't as revolutionary as it sounds. :)

[lupaz]

I don't think it needs to be a default.

Perhaps adding a video to your youtube channel so people that don't follow this person that created the video are aware of these settings and the pros and cons.

I think back when Corona started they used to do a lot more of technical videos that included "secrets" that we didn't know of. Like stuff you could do with certain corona maps that wouldn't even cross our minds, or workflows and techniques that were very interesting and weren't necessarily showing new features only.  Like the way to work with corona scatter and using just as spline to direct a path in the forest.
I think that guy really knew how to use Max in general. Very knowledgeable and was explaining multiple things in one video. Credit where credit is due :)