Lost Worlds

[Seth]

well look!! so much more detail!!

i didn't see the detail 1 version so i can't tell

[Xpleet]

Nice image.


I find it strange that TG's GI makes shadows that are closer-oriented to the sun more light, because in reality it's right the different way, because those shadows have less light that is bounced off and thus become way darker than shadows on the far side, or maybe that's just because it's not using radiosity . Anyway, i see it in almost every Tg2 picture and it's unrealistic.

[Tangled-Universe]

well look!! so much more detail!!

i didn't see the detail 1 version so i can't tell

Exactly, can't really believe this scene needs quality 1.5 or that the difference between detail 1 and 1.5 is that big. If you have very close-up detailed features with micro-displacements etc. then I can imagine it. But ok, enough about that...
I think this is a pretty scene and I like the lighting (well set up) and the colors. Good work!

Martin

[RArcher]

Very nice setup.  I like the lighting and the composition with the planet.  I would suggest that the snow needs to have a greater sense of depth.  You might find that an AA of 25 would be great with vegetation, but for a scene with no models or very small scale details usually I find that an AA of 8 will do just fine.

[dandelO]

Wonderful image... are you sure you didn't hide a plant in there somewhere?  Wink

I like it, Jack! Simple yet effective.

[Matt]

Nice image.


I find it strange that TG's GI makes shadows that are closer-oriented to the sun more light, because in reality it's right the different way, because those shadows have less light that is bounced off and thus become way darker than shadows on the far side, or maybe that's just because it's not using radiosity . Anyway, i see it in almost every Tg2 picture and it's unrealistic.

I think this really depends on the scene, and Terragen simply samples the environment to see where the light is coming from. Typically, if the landscape is brighter than the sky (e.g. a desert lit by a high sun), then most of the light comes from surrounding terrain and will often illuminate the anti-solar facing surfaces more brightly. At dawn/dusk, however, the brightest thing in the scene is the sky, especially near the sun, so sun-facing surfaces are usually brighter even when they are not illuminated by direct sunlight.

Very close to a large object which is blocking parts of the sky, the predominant lighting direction often reverses. If the GI settings are not high enough then Terragen may miss some of that subtlety, but higher GI settings should help. I don't think this image needs it though.

"Radiosity" is just one particular approach to calculating global illumination, which is particularly suited to architecture models, but is not the only correct way

Matt

[Tangled-Universe]

Via via I cam here and interesting remark Matt.

I agree with Xpleet about the strange distribution of lighting in TG scenes.

Recently I had a conversation with a photographer for an art project we are planning to start and we went through a LOT of TG images and his opinion was that all TG images have 1 thing in common: the "reversed" lighting where distant terrain becomes darker instead of lighter as in reality.
I think his observation, and Xpleet's, is correct.

I'm wondering: do you still disagree with this? is it a technical issue that this isn't captured correctly, because of inappropriate GI settings? Or is the way it's done because the lighting model is an interpretation rather than an approximation?

[Matt]

Hi Martin,

I don't really know what you mean. Could you show me an example?

Matt

[Tangled-Universe]

Here, for example:

http://www.flickr.com/photos/66188926@N03/6031706340/in/set-72157627275100045
http://www.flickr.com/photos/66188926@N03/6102275056/in/set-72157627275100045/
http://www.flickr.com/photos/66188926@N03/6028484880/in/set-72157627275100045/
http://tangled-universe.deviantart.com/gallery/#/d3kfur6
http://tangled-universe.deviantart.com/gallery/#/d3l2nk6
http://tangled-universe.deviantart.com/gallery/?offset=24#/d1ywehf

Perhaps not the best examples in terms of the amount of depth, but the effect I describe is noticeable.
It's a kind of 'falloff' of shadow brightness which somehow isn't right.
According to that photographer, who is unfamiliar with 3D, this strange falloff you can notice often is the CG giveaways of many renders.
He describes it at "just by looking at it you feel that there's something off about that treatment of lighting over distance, but I don't know how to describe it other than that it should become less contrasty over distance and in overall brighter".

See a bit what I mean?

Cheers,
Martin

[Matt]

I trust that the photographer is seeing something that's missing from these renders. Atmosphere reduces contrast, but it wouldn't increases the brightness of the surfaces themselves. If that's the effect you're looking for, you need to make a scene where the foreground is more shadowed than the background. Or, add lots of atmosphere to the scene and then crunch the blacks (or gamma down) to make the haze-free foreground appear darker. Those canyon images do seem to have brighter foreground GI than the GI in the distance. In some cases that could be the configuration of the sunlight and the canyon geometry. In other cases, maybe GI relative detail could be a factor, I'm not sure. Remember that many TG users recommend very low GI relative detail, which can miss a lot of detail. Perhaps it would be useful to try brute force monte carlo GI renders of these scenes to see what's missing.

This image:

http://www.flickr.com/photos/66188926@N03/6102275056/in/set-72157627275100045/

I don't think is a good example, because the foreground is in sunlight, generating bounce light off the ground, while the distance is all in shadow. I'm not sure what it says about the problem. But maybe the renderer should have lit the midground cliffs more brightly?

Matt

[jamfull]

Hi Martin,
apologies for jumping in here but, is this one an example of what it should look like or shouldn't?

James

http://tangled-universe.deviantart.com/gallery/?offset=24#/d1ywehf

ps. it's one of my favorites

[Tangled-Universe]

Thanks Matt,

I deliberately chose these examples because I know that Luc, as well as me, always render with GI relative detail at least set at 2.
We're both the patient type of TG guys when it comes to rendering and don't make much compromises since we generally don't mind longer rendertimes.

The example you didn't find very good is understandable for me.
Likely, by the looks of it, Luc used his older canyon setups based on a DEM file (he sold those quite some time ago).
I learned a lot from it, but also that the scales are not real-world. What you see is scale-wise actually 10x bigger.
I don't know what the maximum distance of every bounce is for GI?
Yet, I still found it strange that it became darker in the distance, although from a photographic pov you could expect that since the lighting is in camera.
So yes, probably not a good example.

Do you mean you have a build somewhere with bruteforce MC GI?
As you may know from other recent tests I don't mind rendering a bit longer to test things out.

Martin

[Matt]

Likely, by the looks of it, Luc used his older canyon setups based on a DEM file (he sold those quite some time ago).
I learned a lot from it, but also that the scales are not real-world. What you see is scale-wise actually 10x bigger.
I don't know what the maximum distance of every bounce is for GI?

The maximum distance is 10^16 metres (1e16), which is the same as the maximum distance visible to the Terragen camera. So that's not the cause.

Yet, I still found it strange that it became darker in the distance, although from a photographic pov you could expect that since the lighting is in camera.
So yes, probably not a good example.

Yeah, it might not be correct on these canyon images.

Do you mean you have a build somewhere with bruteforce MC GI?
As you may know from other recent tests I don't mind rendering a bit longer to test things out.

I don't have anything that's working properly yet.

Matt

[Tangled-Universe]

Thanks Matt. Let me know when I can be of any help.

[Matt]

The maximum distance is 10^16 metres (1e16), which is the same as the maximum distance visible to the Terragen camera. So that's not the cause.

Which, BTW, is just over 1 light year