Planetside Software Forums

Support => Terragen Support => Topic started by: pokoy on January 24, 2019, 12:06:59 PM

Title: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: pokoy on January 24, 2019, 12:06:59 PM
I have kept TG 4.2.10 when installing TG 4.3.22.
Today, I confused the shortcut icons and created a scene in TG 4.2.10, saved and closed it, then accidentally re-opened the scene in TG 4.3.22 in order to continue working. When rendering test frames, I realized that the colors were different, tested this and yes - for some resaon, the exact same file comes looks different, see here:

Terragen 4.2.10:

[attachimg=1]


Terragen 4.3.22:

[attachimg=2]

I didn't touch any settings, it's exactly the same file. You can see two things:
1. In 4.2.10, the atmosphere is orange-ish while it's way more purple in 4.3.22, the change is very much noticeable.
2. The ground fog noise comes out a bit differently, but the noise of the mid-level cloud looks identical.

Why is this happening? Whatever the cause is, it really shouldn't happen, or it should be included in the release notes that a certain version 'breaks' the way noises and atmosphere is rendered.
Is there a way to make sure I'm getting 'old-style' renders with TG 4.3.x?
Title: Re: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: pokoy on January 24, 2019, 12:26:44 PM
FWIW, the RTP output in TG 4.3.22 looks way closer to the 4.2 output, but the final render will have that purple tone. I'm puzzled.
Title: Re: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: Matt on January 24, 2019, 02:34:17 PM
Some things do change between versions as I discover bugs and work towards more correct/stable results between adjacent frames. The goal is always to improve quality, correctness or speed (without sacrificing too much of one for another).

There were some changes between 4.2.10 and 4.3.21. Taken from the release notes (change log):

"Solved a problem that caused some unnecessary flickering and reduced the quality of GI in clouds when not using a GI cache file. The new code renders the GI pre-pass faster at low resolution or low detail. High resolution pre-passes may take longer than before, but the quality of the final pass will be improved."

That fix can result in brighter GI in clouds, but might also affect atmosphere, indirectly (although I didn't expect it to be significant enough to log that detail).

"Fixed a couple of bugs in the algorithm that calculates voxel buffer dimensions from the desired number of voxels."

Unlikely to be the cause, but it's possible.

"Removed a bug that caused incorrect cloud dimensions in some animated scenes.".

That last one only affects some animated scenes.

It's possible that one of these changes has introduced a new bug I'm not aware of yet. If so, I'll aim to fix it. But it might just be a consequence of improving the GI solution, in which case you would have to reduce the enviro light amount in the atmosphere if you want to get closer to the previous look.

Can you send me a TGD to test?
Title: Re: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: pokoy on January 24, 2019, 02:47:37 PM
Thank you for the details, Matt. I can workaround this in post for now, but it's good to know about the Enviro parameter should I come across this again.
Attaching is the file, hope you'll be able to find something.

As for the different tone between RTP and final render - I remember seeing this earlier already where the RTP would be slightly warmer than the final output. It makes it hard to correctly preview the final with critical hues when reproducing a real-life shot. Could this be improved in the future?
Title: Re: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: Matt on January 24, 2019, 02:55:12 PM
Thanks, I'll test the file with a few different versions to figure out exactly what change led to the visual difference.

The RTP colour cast difference is difficult to fix because the RTP is designed to provide immediate feedback which restricts the accuracy of the GI (and true brute force GI would be too noisy so we can't use that either). But I will attempt to improve this. I might be able to do this by calculating the accurate GI cache in a separate thread, using the quick RTP solution for immediate feedback and then as soon as the final GI solution is ready the progressive RTP refinement could switch over to that. Not an easy task, but it might be possible to combine it with some other RTP improvements which are coming this year.
Title: Re: 4.2.x vs 4.3.x - render output different hue in atmosphere
Post by: pokoy on January 24, 2019, 03:30:06 PM
Thank you, Matt, sound like a good plan!
I realized recently that the OpenGL preview nicely matches color temperature of the final while RTP does not, that's why I thought maybe it would be possible, especially since RTP is so much faster and really useful. It has added a lot to the workflow, makes working in TG much more enjoyable.