I've looked at clouds, from both sides now....... :-)

You guys... Yes, no one is happy with the current clouds but I'm not sure if buying new clouds is going to solve everything. And yes the trueSKY clouds look a lot better than what we are used to in simulation. But my question with simulated clouds is always: why are people happy with such "bad" looking clouds? This is one of the best examples out there, it looks quite good but not quite realistic, still in the uncanny valley for me a bit... Real clouds are awesome and no sim has really convinced me yet. So I think we can still get better than this, I'm not yet fully happy, hehe

The screenshot below looks good in some areas, for example I really like the horizon on the left and right and the shadows under the build ups in front. I also like that the upper cloud layer seems to cast a shadow on the layer beneath, looks really good. But the edges of the cloud layer at the top look washed out and the transition to the dark background doesn't look as good. Cloud itself should become more white towards the edge there, more washed out and more of a diffuse edge, since that is probably a thick cirrus cloud judging from the monotonous gray underneath. Even cumulus clouds would not have such random parts in front of the clouds, there is usually a sharp edge or a diffuse wash out, never sharp individual blobs like seen here. I also don't like the cloud structure of the cloud "blobs" in the lower level (smaller build ups), they looks too random. Clouds are not random, they have structure where they are build and they fall and dissolve on the other end... This looks more like a huge propeller chopped the clouds into pieces. It's the fractures of the clouds that are too large in size yet, should be much finer, much more homogeneous, less spot like if they are that transparent.

[Blocked Image: https://cdn.80.lv/80.lv/uploads/2017/08/4.2.png]

This one below doesn't look good I think, the light doesn't match the mountains in the distance at all. And again, too little structure compared to real worlds.

[Blocked Image: https://cdn.80.lv/80.lv/uploads/2017/08/FutureLighthouse-UE4-Tomorrow_VRShort.jpg]

Here, take this how it should look like:

Take these time stamps:

1) 0:16 - milky soup below, topped off with some highly fractured clouds, fragments are relatively local, there are typical cauliflower, cumulus like, cloud tops, piles of clouds towards the left side. Then in the higher altitudes you have a giant cirrus, very soft and unstructured cloud and underneath that some cloud fragments, small in height and relatively stretched along lines... And also look at the lower edge of the upper layer, really soft with some fragments hanging lower.

2) 0:19 (and 0:23) - very wavy structure in the lower left corner, dense clouds with sharp cloud tops, embedded in a lower level haze layer. Sharp clouds are faded out towards the edges, transition into the haze, really. In the far distance: only mountain tops peaking out of the clouds, scenery invisible underneath. Really soft transition of the low haze and the mountains

4) 0:27 - this is what I want to see in a next gen sim. That looks insane. OK, we got actual 3D looking cirrus clouds with really soft edges and feather like structure (not blobs as seen in the screenshots above), medium high level clouds with quite defined edges (feather like or soft drop off line, again no thick blobs at the edge), large wave like structures in the left cloud layer. On the right hand side, another smaller wave like cloud (mix of cirrus, lenti cularis and alto cumulus?), that one as more broken off edges, but the height of these fragments taper off towards the edge as well, really thin border, lets a lot of light through -> brightens up when seen from below, not visible in the sceenshots above. An then of course high altitude build ups in the lower right corner, nice cauliflower structure with fine fragments in between. And finally the beautiful towering cumulus build ups in the bottom center, illuminated from the back, edges very bright, core very dark. Overall shape looks like a mountain, higher in the middle, tapers off towards the edge,

5) 0:40 Cumulus with sharp tops, soft and diffuse fragments in between, clouds are very local

6) 1:08 transitioning through the haze into the overcast layer, note how soft the haze is, the view gets worse but you can still see very far towards the upper clouds, visibility down into the haze is really bad. Also a thing we don't see in simulation yet, different visibility range depending on where you look at... Also, you can't see the city below but from below, looking up, you can see the upper cloud layer

7) 1:27 (that rain shower! :3 aw man I want that! Clearly marked edge of the rain shower, very hazy looking towards the upper cloud layers, nicely indirect lighting of the entire scene, and small raindrops, then larger ones,... soaked wet runway and apron with high reflectivity, almost like a mirror, yeaaaah I want that

2:09 a lot of locally increased moisture with "almost clouds", some thermal tops reach high enough to form a cumulus cloud

9) 2:14, if we ever get a lighting like this in the sim we made it Purely indirect light through multiple cloud layers

Now, this is all insanely difficult to program, I can imagine, but we shouldn't stop demanding better clouds. Again, I personally, am not happy with any clouds out there presented as highly realistic. And I am getting a bit tired when people scream: look at those clouds, they look fantastic, when they simply aren't. Yes, they may look better than the default clouds in this or that simulator but I'd still only rate them 5 out of 10. I'd rate the first screenshot of those trueSKY clouds an 8/10 (which is really good for a real time rendering) and the second picture only a 3/10. I think Aerofly deserves clouds that look 8 or 9 out of 10 and 7/10 in the worst conditions. The screenshots shown for trueSKY don't quite convince me, as I said, it could still be a lot better. "Good but not quite good enough" I think.

Like we have previously said, we plan to improve on our clouds, but the real
issue with a 3rd party solution is the integration. As we have previously seen

in the past, any 3rd party solution adds a tremendous overhead when it comes
to customising it and in case of clouds we need very fine grained control to
match it with our wind and atmospheric engine.

We are not saying 3rd party solutions are bad, the clouds look good, but we
then have to spend a lot of time to integrate it into our engine and each new
platform we work on, requires additional work due to this 3rd party addon.

One of the reasons why Aerofly runs so smooth is, that we have coded all
the stuff ourself, so its much easier for us to adapt to new platforms, enhance
things or switch over to other engines.

Quote from Mr Crowley

Has IPACS looked into procedural generation for volumetric cloud creation?

And could it actually work in the AF2 engine?

At the moment there's a lot of work yet to be done with other features. We will get to the weather at some point but don't expect it any time soon.

Quote from ussiowa

Ideally the computing power should be used in a sphere around oneself, so it's either used for cloud or terrain, whatever is visible around without obstruction.

We can get more efficient than that... You don't want to render clouds or scenery behind you, waste of computing power.

They still have to be loaded and their position has to be calculated etc, so if you turn your camera that they don't have to load in and create lag or visual artifacts.

Aerofly scenery is only rendered in front of the camera I think, but I'm not 100% sure. Ideally you would only render a section of a sphere, only the direction you are looking at.

Quote from Almdudler

Can you do it? Can you build a next generation weather model?

at a reasonable time tag?

If you believe you can, please go ahead and do the bloody best thing ever done for simulation.

Marc (spelled with a c!) and Torsten are two very smart guys. So yes they can do it and in a reasonable time tag probably yes. I mean they have already created the Aerofly engine all on their own and you can test the excellent result for yourself. If they can do that they are probably also capable of "rendering some clouds and animating them"...

I have big hopes for a next generation weather engine and I hope I can be involved in the development as well because I have a bit of experience with real world weather, thermals and mountain waves, slope soaring, etc.

The question is rather how many features will be included. Do you need historical weather for each and every minute of the last 100 years, what about rain, snow fall, snow on the ground?, hail that damages your airplane? ice build up? what about turbulence that is strong enough to damage your airplane as well, how about wind shear, wake of other aircraft? mountain waves, rotor clouds,... the list goes on and on. That all takes a bit of time, so how far is good enough?

Personally I'd already be happy to see an overcast sky with local rain showers with the occasional lightning and thunder (so rare that is surprises me), wet runways and increased turbulence in those showers, clear skies or only some cirrus clouds once I break through the ceiling. Or a fair day with cumulus clouds and thermals with slight to moderate turbulence underneath, just enough that you want to slow down in your A320 to not get thrown around by the turbulence, some wind gusts due to thermals, slight average wind, etc.