New screenshots, still no cirrus clouds

[mp15 138]

@Weatherman did you like my cloud design?

[mp15 138]

@Weatherman did you like my cloud design?

[JAS8472 10]

Hey mp15, you make the worlds greatest designs (or even in the universe), but let’s focus on finding or discussing about the implementation of cirrus clouds in the upcoming MS Flight Simulator. 

[Dominique_K 10,138]

1 hour ago, Weatherman said:

 But this leads into too much meteorological detail here.
All My guess is that is is not easy to build these with the volumetric approach Asobo's engine takes. B 🙂

Thanks. Don’t think thats too much details, this forum is a good place to learn things !

Why Asobo’s volumetric approach would not lend itself to simulating cirrus ?  They  are a volume of crystals after all ? 

From wikipedia: 

Cirrus clouds range in thickness from 100 m (330 ft) to 8,000 m (26,000 ft), with an average thickness of 1,500 m (4,900 ft). 

[Weatherman 193]

I am not a graphics engine expert, so to be honest my guess is not based on any expertise. It's just my feeling, that something like cirrus, which in most cases has a very high transparency by it's nature and still looking very uniquely defined and detailed from the distance at the same time might be hard to build by volumetric modeling. At least if they try to achieve the same level of visual quality as for the more cumulus style cloud types we have seen so far. But I could be completely wrong. Does anybody know about some other engine that has the ability to produce good looking (volumetric!) cirrus?

[Dominique_K 10,138]

A video of a MAF VFR flight in PNG.  Several sequences with clouds. Asobo clouds are not far. I just wish they were a tad less « spectacular ».

 

[Skip Talbot 17]

On 5/10/2020 at 12:18 PM, Weatherman said:

There have been some screenshots with higher altitude cloud layers that might be supposed to be cirrus but to me they resemble more altostratus or altocumulus layers. 

 

14 hours ago, Dominique_K said:

I am not a specialist but don’t we see cirrus at the 3’03’’, 3’10’’ and 6’13’’ marks ?

Yep, I think that's what they're intending to be cirrus. Folks are asking for the fine filament structure though, not just a thin layer of high altitude generic cloud material. 

 

 

10 hours ago, Weatherman said:

I am not a graphics engine expert, so to be honest my guess is not based on any expertise. It's just my feeling, that something like cirrus, which in most cases has a very high transparency by it's nature and still looking very uniquely defined and detailed from the distance at the same time might be hard to build by volumetric modeling. At least if they try to achieve the same level of visual quality as for the more cumulus style cloud types we have seen so far. But I could be completely wrong. Does anybody know about some other engine that has the ability to produce good looking (volumetric!) cirrus?

Most "good looking" cirrus, or any high level clouds with a distinctive pattern like mackerel sky or mares tail, that I've seen is implemented by hand drawing or using photos on a flat texture. These are like "fake backdrops" in games where interacting with the weather isn't important. Obviously flat textures aren't great for realism reasons, and they're difficult to animate realistically. 

When making volumetric clouds, it's common practice to combine different scales of 3D noise together. Here's 3D noise contained inside of a box:

 

You start with big blobs of noise. And then you make smaller noise and add or substract the smaller blobs to the big ones. Make the edges fuzzy by varying the density, and scatter the light inside of them, and you've got a good looking cloud. Flight Simulator cloud:

 

If you watch episode 7 of the feature discovery series, there's a shot where they timelapse traffic at an airport. But look at the clouds overhead. It gives a great glimpse as to how they're making these clouds. The "large blobs" of noise grow and shrink slowly. And then it looks like they're taking that box of "smaller blobs" (see first shot of 3D noise) and they're simply moving the box in the vertical direction, while keeping the larger blob base cloud in place. So that small blob noise pattern is translating up making it look like the cloud has an updraft. It's simple but effective, and in realtime it's a nice way to make 3D clouds smoothly animate and transition. But sped up, it looks a little unrealistic (to a cloud geek like me at least) as cumulus grows more like a tree along branches and decays along those branches as well. 

 

A peek at the weather controls from the same video: 

 

Looking at the variables you get to control there, I'm guessing all the cloud "types" are constructed along similar lines. They're variations of different scales of noise. They'll all look kind of blobby as a result. It might be too computationally expensive or complex beyond the scope of the game to implement a more realistic model of how clouds form. The simulation of the supercell with tornado I linked to earlier was created with one of the fastest supercomputers in the world and generated several hundred terabytes of data. We might have to wait until FS2040 to get that in game. There may be more advanced procedural patterns they could use other than noise that look better for cloud types like cirrus or stratocumulus, but even that might be expensive on the GPU or out of scope for the developers. 

Even so, we might be able to see the thin and delicate wisps of cirrus using a 3D noise pattern. Instead of big blobs of noise added to small blobs of noise, you'd start with just the small blobs. Make each blob very small and spaced far apart, squashed into a more flat layer. Then you stretch those blobs horizontally like taffy (extrude them). It would probably require a lot of tweaking to make it look decent, but you'd probably get some nice volumetric streaks. They'd then animate and transition in the same fashion as the other cloud types. 

Right now it looks like we've just got high altitude pancakes to work with, meaning they made big blobs and just flattened them. 

Edited May 12, 2020 by Skip Talbot

[MaVe64 414]

This is what I hope to see. Watch and enjoy!

 

 

 

 

[dtrjones 356]

On 5/11/2020 at 1:06 AM, mp15 said:

@Weatherman those clouds could be drawn in Photoshop by a 4 year old, Asobo has volumétric cloud creations technology, so expect to have those simple 2D clouds in game too.

Isn't that the issue, to have great looking cirrus clouds they need to be 2D textures, they don't render well in 3D, why isn't anybody mentioning this? Are you all chicken! 😃

I can't see a reason why Asobo can't mix 2D with 3D for an effective cloudscape, maybe this one has to wait for third party devs.

[dtrjones 356]

11 hours ago, Dominique_K said:

Thanks. Don’t think thats too much details, this forum is a good place to learn things !

Why Asobo’s volumetric approach would not lend itself to simulating cirrus ?  They  are a volume of crystals after all ? 

From wikipedia: 

Cirrus clouds range in thickness from 100 m (330 ft) to 8,000 m (26,000 ft), with an average thickness of 1,500 m (4,900 ft). 

I suspect it's difficult to render a cirrus cloud in 3D and make it look good viewing from ground to through to 35K feet. Hopefully Asobo will crack the nut and we'll have great looking cirrus clouds, as you say they are 3D after all. 🤞

[scotchegg 4,186]

1 hour ago, Skip Talbot said:

 

When making volumetric clouds, it's common practice to combine different scales of 3D noise together. Here's 3D noise contained inside of a box:

 

Sorry for the off topic, but that looks very much like models of gluon fields:
 

 

[avhpilot 156]

Perhaps the lack of Cirrus is for the same reason X-Enviro cannot do it in X-plane

https://forums.x-plane.org/index.php?/forums/topic/123418-no-clouds-at-high-alitude/