Sombrero

Started by WAS, April 08, 2018, 04:56:24 PM

Previous topic - Next topic

Denis Sirenko

An interesting thing. I also made a generator of galaxies for my purposes (for our game), but it was not good as yours. To accelerate the rendering, did you try to manage one single cloud layer v3?

WAS

Quote from: Denis Sirenko on April 13, 2018, 03:33:49 PM
An interesting thing. I also made a generator of galaxies for my purposes (for our game), but it was not good as yours. To accelerate the rendering, did you try to manage one single cloud layer v3?

I tried several editions with as few cloud forms as I could, my current project has 5 but still a bit slow. Also figured out using voronoi diff scalar produces cross banding seen in a lot of galaxies/nebula. Just getting them to play nice and look natural is my current issue.

Denis Sirenko

#17
Quote from: WASasquatch on April 14, 2018, 03:41:04 PM
Quote from: Denis Sirenko on April 13, 2018, 03:33:49 PM
To accelerate the rendering, did you try to manage one single cloud layer v3?
I tried several editions with as few cloud forms as I could, my current project has 5 but still a bit slow.

Perhaps I have inaccurately formulated the question. Did you try using one single cloud layer? Not 5, but only one for all the necessary components of galaxy. First, combine the fractals of all the necessary components (through the merge shader), and then send it to the cloud layer's density input. To paint such a nebula will be necessary through a direct light modulator.

And I have a question on the image:

[attach=1]

Do these stars exist in 3D space or is it a background? Did I understand correctly that you used this starry background — https://planetside.co.uk/forums/index.php/topic,19264.0.html — to connect it to density input of Cloud Layers?

WAS

#18
Quote from: Denis Sirenko on April 20, 2018, 07:26:13 AM
Quote from: WASasquatch on April 14, 2018, 03:41:04 PM
Quote from: Denis Sirenko on April 13, 2018, 03:33:49 PM
To accelerate the rendering, did you try to manage one single cloud layer v3?
I tried several editions with as few cloud forms as I could, my current project has 5 but still a bit slow.

Perhaps I have inaccurately formulated the question. Did you try using one single cloud layer? Not 5, but only one for all the necessary components of galaxy. First, combine the fractals of all the necessary components (through the merge shader), and then send it to the cloud layer's density input. To paint such a nebula will be necessary through a direct light modulator.

I actually have no tried merging all the cloud forms into a single layer, and I do wonder how this would impact renders. I'm not sure if I'd be able to modulate the depth appropriately as it uses multiple clouds for different depths, but going to give it a try since I haven't even added any depth to this new setup I am trying.

Quote from: Denis Sirenko on April 20, 2018, 07:26:13 AM
Do these stars exist in 3D space or is it a background? Did I understand correctly that you used this starry background — https://planetside.co.uk/forums/index.php/topic,19264.0.html — to connect it to density input of Cloud Layers?

These were isolated from the Glitterato plugin from Photoshop and warped to be on a "plane" and just masked into the galaxy. The starfield I have not actually played with. I know in order for it to work you'll need a spotlight behind your nebula seen, pointing in the same direction as your camera, with a wide focal, and super bright settings. Because the background has no luminosity, you'll need to provide it light (it gets light from the Sun).

Alternately you could use a default shader and a PF and luminosity I theorize, though have never attempted. I think I may play around with those today. Been slowing down on work just due to being so hindered on a 2x5 inch System on a Chip computer haha.

Edit: I figured out the star background IS luminous. It just uses a distance camera as a mask which doesn't quite work for our setups. Though having a issue with what is produced via rendering.

Denis Sirenko

#19
I'm understood, thank you.

As for the use of a single cloud layer. If you manage to figure out the heights and set up a single farctal on the density input of a single cloud layer - you can save a lot on render time. My current nebulae are realized in one cloud layer.

As for the stars. I realized the luminous stars in the 3D TG space. They are rendered in TG at the same time as the nebula. They can be an overlapping nebula in a natural way.

[attachimg=1]

For this, in fact, it's enough to send a fractal, similar to this (https://planetside.co.uk/forums/index.php/topic,19264.0.html) to the density input of cloud layer v3. But I used the Voronoy to make the stars round. The luminescence is achieved with the help of the ambient input of the cloud layer.

Moreover, it is possible to achieve that the nebula itself is illuminated only by these stars! Above the nebula is illuminated by light sources. And so it looks like a nebula illuminated only by procedural stars:

[attachimg=2]

But there are two problems:
1) Each new rendering is a new stars. This is corrected by increasing the size of stars or increasing the setting of AA
2) The number of stars increases near the nebula. And the reason is not that the fractal for cloud adds color to the fractal of stars. I think this is due to an increase in the accuracy of calculation around the nebula.

WAS

Wow, this is a nifty approach, I'll have to fiddle with it. I wonder if the AA limitation of 6 I have will effect results? Still, overall that's great work. I remember I tried a similar setup to try snow.

WAS

#21
Wow seems really promising. Liking the effect.

A issue I do have with my setup is there are seams between tiles. You can see some stars cut-off. And interestingly, adding more AA detail, creates more stars. AA6 compares to AA2 has waaaay more stars.

Denis Sirenko

Quote from: WASasquatch on April 22, 2018, 12:00:23 AM
Wow seems really promising. Liking the effect.
A issue I do have with my setup is there are seams between tiles. You can see some stars cut-off.

I have faced this recently, when I researched the work of Fractal warp shader. I noticed a tendency to exhibit tiles. But here seems to be no need for a warper. Now I do not find this tiles, I use only Power fractal shader v3 (Voronoy) + Cloud layer v3.

I suspect that it can also be in the render node settings. Did you notice, are these joints on the borders of threads that render visualization?

Quote from: WASasquatch on April 22, 2018, 12:00:23 AM
And interestingly, adding more AA detail, creates more stars. AA6 compares to AA2 has waaaay more stars.

I think this is somehow related to the probability of a tracer beam falling into each particular star during rendering. Pay attention that the brightest stars are more likely to re-meet on the render than the weak stars (with each rendering of the same scene).

In general, I'm studying the question of creating stars in order to include this in animation and for me this behavior can be a problem.

WAS

#23
Quote from: Denis Sirenko on April 22, 2018, 03:25:29 AM
I suspect that it can also be in the render node settings. Did you notice, are these joints on the borders of threads that render visualization?
Yeah, all the clipping is happening against thread borders. I assume it's how each cell is calculated, it assumes "Hey, there's a bright star here" while the next cells interpretation says "Nope, nothing there".
Quote from: Denis Sirenko on April 22, 2018, 03:25:29 AM
In general, I'm studying the question of creating stars in order to include this in animation and for me this behavior can be a problem.
Yeah, I believe this would create stars out of no where as movement is created in a scene, which is disappointing. But I am wondering if we can just use a basic repeatable noise map from Photoshop or something to govern stars. They would be "fixed" and not subject to so much clamping calculation within TG, perhaps this would define a stars position more "definitively".

Here is a simple test throwing the stars setup into a thin galaxy bands cloud layer. No galactic core light source, this is all dependent on stars light emission. I'm still not totally liking my star colours, but man did it take ages of testing to ensure there were no green/purple erroneous stars. Colours interact weird when they gradient in TG. Now just need to get balance between star types. Should be more red dwarfs and main sequence yellow stars. Blue and White are already pretty rare, but should be rarer. I would have just use the Night Sky's color system, but it unfortunately favors unrealistic star colours.

Added another test with the core active and more luminous stars.

Oshyan

QuoteI think this is somehow related to the probability of a tracer beam falling into each particular star during rendering. Pay attention that the brightest stars are more likely to re-meet on the render than the weak stars (with each rendering of the same scene).

Yes, I think this is pretty much the problem here. But due - I think - to the "stars", i.e. the bright points, being too small, too noisy. So try to make your "stars" a bit larger and less of them, less dense, if possible. You need to make it so it's easier for the renderer to consistently and effectively sample your star function.

- Oshyan

WAS

#25
Quote from: Oshyan on April 22, 2018, 08:24:24 PM
QuoteI think this is somehow related to the probability of a tracer beam falling into each particular star during rendering. Pay attention that the brightest stars are more likely to re-meet on the render than the weak stars (with each rendering of the same scene).

Yes, I think this is pretty much the problem here. But due - I think - to the "stars", i.e. the bright points, being too small, too noisy. So try to make your "stars" a bit larger and less of them, less dense, if possible. You need to make it so it's easier for the renderer to consistently and effectively sample your star function.

- Oshyan

I tried to give this a test and it didn't seem to make much of a difference. Here is a difference comparison between the two images. As you can see there are new stars, and some stars have shifted slightly (green/blue)

I'm testing a new render using a larger minimum scale to try and rid the smaller stars. You can see the preview in Screenshopt_29.png

WAS

Seems to be just a cloud limitation issue. Even huge stars do the same thing. Even at about 10th a km, they will disappear at random distance or more pop in close, and same issue with AA.

Denis Sirenko

It seems, this is only suitable for a static scene, when randomness is not terrible.

Oshyan

I'm honestly not sure how your cloud-based star method works, so I can't say exactly what the issue might be. I guess your clouds are luminous though?

- Oshyan

Denis Sirenko

Oshyan, I translated the discussion of these stars here: https://planetside.co.uk/forums/index.php/topic,24396.0.html.
I think we deviated from the original topic theme, which created WASasquatch.