Why it is hard to add clouds in the Sky?
- Thread starter Ratstar
- Start date
The visual effect of clouds is not the trouble. It's the effects of an atmosphere on the flight characteristics of a space craft. Just like you get the effects of gravity on your ship as you approach the surface now, go ahead and add the variables for atmospheric re-entry, and controlled flight to the whole thing. To do it the way the ED of old would, and I don't disagree, they wouldn't want to add that kind of approach, without the simulated affects of wind, density change, and the like on your craft. You know, game-plsy.
The ED that has been displayed recently just might. I would hate to see anything that mars the flight dynamics in Elite. They are unique and compelling. But, FDev brought us the SCO drive, which successfully maintained the scale of these star systems, but also made SC way more engaging. Maybe they could find a path, who knows?
The ED that has been displayed recently just might. I would hate to see anything that mars the flight dynamics in Elite. They are unique and compelling. But, FDev brought us the SCO drive, which successfully maintained the scale of these star systems, but also made SC way more engaging. Maybe they could find a path, who knows?
What sort of cloud did you want? Cirrus, cirrocumulus, cirrostratus, altocumulus, altostratus, nimbostratus, cumulonimbus, cumulus, stratus or stratocumulus? Maybe some exotics like notilucents and lenticulars?
The obvious problem with planetary clouds is multiplayer synchronisation.
- the clouds need to show the same for every player in an instance (you think you're hidden by fog, but your friend has a bright and clear day)
- the clouds need to have believably consistent structure from orbit-to-surface in the same way that the surface terrain does
So far, that's all things Frontier has already done, or close analogues of it. Surface terrain and the asteroids in ring systems are good examples.
The tricky bit is if the clouds are supposed to move (and it'd be weird if they didn't). Then the problem goes from "generate a realistic-looking cloudscape" to "generate an infinite number of realistic-looking cloudscapes in a time series where each time step is believable, and be able to produce any of those cloudscapes without iterating from the previous time steps"
This is something Frontier has done at a much simpler level - the orbits of planets in a system behave this way, and most players aren't sufficient experts in astrophysics to notice during normal gameplay the inevitable simplifications it introduces. Asteroids in ring systems spin on the spot, though very importantly don't move relative to each other. Whether the techniques used there can be generalised to faking up entire planet-wide weather systems? I don't know - and importantly, for deciding whether to commit time to it or do something else, probably neither does Frontier yet.
- the clouds need to show the same for every player in an instance (you think you're hidden by fog, but your friend has a bright and clear day)
- the clouds need to have believably consistent structure from orbit-to-surface in the same way that the surface terrain does
So far, that's all things Frontier has already done, or close analogues of it. Surface terrain and the asteroids in ring systems are good examples.
The tricky bit is if the clouds are supposed to move (and it'd be weird if they didn't). Then the problem goes from "generate a realistic-looking cloudscape" to "generate an infinite number of realistic-looking cloudscapes in a time series where each time step is believable, and be able to produce any of those cloudscapes without iterating from the previous time steps"
This is something Frontier has done at a much simpler level - the orbits of planets in a system behave this way, and most players aren't sufficient experts in astrophysics to notice during normal gameplay the inevitable simplifications it introduces. Asteroids in ring systems spin on the spot, though very importantly don't move relative to each other. Whether the techniques used there can be generalised to faking up entire planet-wide weather systems? I don't know - and importantly, for deciding whether to commit time to it or do something else, probably neither does Frontier yet.
The multiplayer aspect should be handleable by the proc gen - same as surface features. After all, in a multi-player instance, every SRV will stumble (and occasionally explode) on the same rocks and encounter the same brain trees.
Lots of multiplayer games have clouds. ED needs more variety for the different atmospheres though. Fdev just hasn't prioritized new planet types with could tech.
What a silly question! All of them!
rootsrat
Volunteer Moderator
Interestingly, in 3.8 and earlier, you could add more small rocks on the surface by manually editing their multiplier in the graphics configuration file.
The extra rocks were not physical (SRV was clipping through them) and they were obviously not visible by other players, as thery were generated at client level.
Last edited:
But as I point out in the following sentences of my post, rocks don't move.
There are only two procedurally generated things in the game which move:
- planets in their orbits
- rocks in ring systems spin on the spot
Both of those are possible to calculate the position and rotation of at any arbitrary time with a very simple set of functions, such that its position/rotation at time X+1 is believably continuous and consistent with its position/rotation at time X, but you don't need to run through calculations from time 0 to time X to find out where it is at time X+1.
Is that possible to generalise all the way to a believable-to-a-non-expert planetary weather system where clouds can form, merge, dissolve, move, etc as you watch them out of the window of your ship? If it is, I'm not aware of anyone else having done it already.
Ok - we might be up against the difference between "theoretically possible" and "practically feasible". Theoretically, a detailed weather simulation is possible. Practically, our best existing computers are not up to the task and are limited to rather coarse grids.
And yes, in practice ED isn't even up to correctly simulating lights and shadows - especially not with multiple light sources (stars) in a system.
And yes, in practice ED isn't even up to correctly simulating lights and shadows - especially not with multiple light sources (stars) in a system.
I play iL-2 Sturmovik which has very nice fluffy clouds of different types - they move depending on the wind conditions - and appear the same to all players in a multiplayer instance. Same happened when I was playing Warthunder. They look really nice in VR and are always fun to fly through to get droplets on the canopy. There are set cloud models (as far as I can see, I’ve noticed an “anvil top” style cloud a few times which is pretty distinctive) but the way they are mixed leads to pretty unique looking vistas.
I don’t understand the argument against clouds based on them having to be ultra realistically modelled. The cloudy planets at the moment are just a clever texture, but is it beyond possibility that you could use that as a LOD basis? Assign different volumetric cloud types to different areas of the texture, fade them in as you get close enough.
If someone wants to sit on a planet for hours and then complain that the clouds haven’t changed realistically, then let them fill their boots (have they done this for the current Earth-likes?). If it looks good for the amount of time currently spent flying down to planets and doing missions etc, that’s good enough for me.
I don’t understand the argument against clouds based on them having to be ultra realistically modelled. The cloudy planets at the moment are just a clever texture, but is it beyond possibility that you could use that as a LOD basis? Assign different volumetric cloud types to different areas of the texture, fade them in as you get close enough.
If someone wants to sit on a planet for hours and then complain that the clouds haven’t changed realistically, then let them fill their boots (have they done this for the current Earth-likes?). If it looks good for the amount of time currently spent flying down to planets and doing missions etc, that’s good enough for me.
True - but there is significant room to cheat in Elite Dangerous because most players aren't exo-meteorologists. It doesn't have to be a perfect simulation or even all that realistic - it just needs to produce clouds which look a bit like clouds without "It was raining when I docked at Noir Depot. But then, it was always raining at Noir Depot." situations.
I think it's analogous to the difference between the ways Elite Dangerous handles artificial gravity and which ones get complaints:
- outposts: supposed to have zero gravity, rubbish sticking to the floor is obviously weird even to a non-expert
- coriolis: rotational pseudo-gravity doesn't work as depicted in the concourses, but most of us don't have the physics background to realise that and most of those of us who do don't work with rotational pseudo-gravity anywhere near enough to be able to tell it's wrong just by looking
- spaceships: very limited interior display detail and some technobabble handwaves enough that we mostly don't notice that all our passengers should be dead after the first time we do a boost-turn, and screaming at us in terror over the intercom even in more routine flight.
Once they start adding clouds, they need to be better than the outpost standard (obviously wrong to a non-expert) and the spaceships standard (fine so long as it mostly stays off-screen) ... but massive simplifications from reality like the coriolis one will pass generally unnoticed.
Something like procgen a layer of clouds/weather that moves independent of the surface?
This is a pointless and somewhat dishonest conversation chain. ED supports sky boxes and sky shaders. The flight model isn't the reason you can't land on more planets. Procedurally generating something of actual value that's worth enabling the planets is the issue and it's going to be more work than all of Odyssey combined to do.
Well they can draw clouds so we should get more planets is never going to float. They'd be worthless if slightly prettier and then you're just going down the full daisy chain of feature creep to what OP most likely actually wants which I'm pretty sure OP knows is far too much work. So we get this implied they can do something easy that's not the issue so give me a whole feature and if OP doesn't actually understand the scope of making atmospheric planets worth landing on. I'd suggest sitting down and just writing out a list of things required to make it worthwhile and not just more brown dead rocks.
Well they can draw clouds so we should get more planets is never going to float. They'd be worthless if slightly prettier and then you're just going down the full daisy chain of feature creep to what OP most likely actually wants which I'm pretty sure OP knows is far too much work. So we get this implied they can do something easy that's not the issue so give me a whole feature and if OP doesn't actually understand the scope of making atmospheric planets worth landing on. I'd suggest sitting down and just writing out a list of things required to make it worthwhile and not just more brown dead rocks.
Yeah, in older ED versions you could manually increase the level of detail on planets and the textures of un-accessible planets. Also more dust, particles, draw distance. It looked better than the highest in-game settings.
Last edited:
Mammatus clouds. Because of course
This used to be a big problem in DCS and was only resolved a few years ago (IIRC). For context, DCS was released in around 2012, so even older than E: D and it took the devs more than 10 years to solve that particular problem
Ooh lenticulars please, they're my favourite!
I prefer Morning Glories.Mammatus clouds. Because of course
Stick a few of these up.
Nobody will really know the difference and the devs could be in the pub by lunchtime. Everyone wins.
Nobody will really know the difference and the devs could be in the pub by lunchtime. Everyone wins.