[1.1.2][1-1-2] May 13-2016 EnvironmentalVisualEnhancements

[jrandom]

I just now tried making some higher-res cloud maps from the huge originals from NASA's Blue Marble site, but KSP crashes (and not because it hit its memory ceiling either) -- apparently Photoshop CS5 does not know how to properly write out 32-bit .tga files. Later I'll see if I can run the full-res original (40K) or a scaled down high-res (20K) through Paint.net and see if I can generate a good 8K map from that.

[TMS]

Just been using 5.3. First time I've tried this mod, having seen it on scott manleys new series.

Having a weird issue where despite being well above cloud cover in orbit, the ship appears to disappear beneath the clouds when I zoom out the external view.

I'm running kethane and universe replacer. Anyone else experiencing this?

[rbray89]

Just been using 5.3. First time I've tried this mod, having seen it on scott manleys new series.

Having a weird issue where despite being well above cloud cover in orbit, the ship appears to disappear beneath the clouds when I zoom out the external view.

I'm running kethane and universe replacer. Anyone else experiencing this?

That makes sense... The camera rendering the overlays exists between the near and far camera... so if you zoom out enough to have something on the far camera, it could appear behind the clouds.

[Nick27222]

This looks beautiful!

[nothke]

All the tips and tricks about rendering volumetric clouds fast is here:

http://ofb.net/~niniane/clouds-jgt.pdf

..maybe visit modders irc more often? We had a discussion on the paper recently =) But didn't know you were "into" volumetric clouds. Several modders were trying some stuff.

[Zander]

Downloaded for the volumetric clouds and overhead cloud layer. No readme, no instructions of any kind in the download. just opened the GUI and started doing stuff. I managed to create one volumetric cloud by clicking generate test cloud over launchpad and I created a cloud layer in the sky but it also made the entire planet white when viewed from space.

This definitely needs some instructions included. The volumetric cloud I managed to create did look very good though.

What I would like, and I think most people would like, is random weather. maybe some days overcast and some days clear skys with a few volumetric clouds. random cloud ceiling heights. just have a random weather profile created each time I launch.

[nothke]

I think the one you generated is the only one.. It's a test cloud. =) The other buttons just edit the cloud layers and not the volumetric cloud.

Btw, implementing weather is not as easy as it sounds. But it would be really awesome. ..With thunderstorms and lightnings hitting your huge rocket, that is xD

[Kalista]

I think the one you generated is the only one.. It's a test cloud. =) The other buttons just edit the cloud layers and not the volumetric cloud.

Btw, implementing weather is not as easy as it sounds. But it would be really awesome. ..With thunderstorms and lightnings hitting your huge rocket, that is xD

Or attempting to land a space plane/shuttle in Kerbins version of a hurricane hitting KSP. Heck more I think about it more I want to do exactly that.

[Zander]

Or attempting to land a space plane/shuttle in Kerbins version of a hurricane hitting KSP. Heck more I think about it more I want to do exactly that.

Of course, random weather would be awesome. it could be clear skys when you launch your spaceplane but a few days later when landing it a heavy storm. and you could get weather reports at the KSC and have to make changes to your mission based on weather conditions or if brave risk a dangerous landing or launch in high winds and low visibility.

[AdmiralTigerclaw]

The problem with something like weather is the computation intensive requirements for a 'weather engine'.

Games like Minecraft pull off weather by just applying a global condition. Since nobody can ever see 'the entire world' in games like that, it doesn't matter what the weather is doing from place to place.

But in KSP, you have to realize that applying a 'global event' to the weather, like a world wide storm, gets pretty silly. Thus, you have to break kerbin into areas. Flight Sim X does this and has sectors in which whatever meteorlogical conditions are present, are set for the entire area. While it makes weather doable, the weather is also 'robotic' for a lack of a better term. You fly from one sector into another and it's like hitting a brick wall as winds instantly change direction and speed, and conditions shift from 'decent' to 'horrid' in an eyeblink. And the conditions only affect you when you reach them, so you don't see them ahead of you either. (In real life, you can see rain from a storm ahead of you.)

Ideally, weather would be broken down, not based on the surface, but into moving weather cells. Each cell containing a 'system'. High pressure or low pressure. High pressure systems would be sunny with patches of fluffy clouds and high cirrus, and Low pressure systems would contain various types of inclement weather. Rain, storms and so on.

But even then, you run into the problem of these cells being static in themselves. You see the low pressure area, and it's the same shape and size just moving around the planet. Kind of blah. You have to break these cells down further into sections containing weather patterns such as cold fronts, dry lines, and other meteorlogical disturbances. And at the risk of sounding redundant, you have to break THOSE down into storm cells in order to truly make the weather non-static.

By the time you get down to 'local cell' events, you've stacked at least three to four tiers interconnected patterns on top of each other. Each section of each tier breaking into dozens in the tier below. By the time you get down to the individual local cells of the storm you're flying through, the computer is busy handling hundreds of these local weather cells. And these cells HAVE to relate to each other or you face the brick wall issue and unrealistic patterns. (Winds that blow opposing directions when they shouldn't, rain where rain shouldn't be... that kind of thing.)

Without some miracle of an optimization algorithm, you're looking at a CPU melting weather simulation system ON TOP of a spaceflight physics simulator that's already pushing it. There's a REASON NWS weathercasting uses some of the most powerful supercomputers in the world for forecast modeling. Even a simplified weather handling system would make the best desktop PC cry.

[Aknar]

I find clouds to be too dispersed and transparent so i'd be happy to get instructions on how to make it look denser and more grouped rather that fogging up the space view. The volumetric cloud looks really great. It must use too much resources to put them all over the place but if they decrease in volume depending on distance, it could be feasible couldn't it? I'm really impatient to see this. It would really add to the environment and immersion to go through the clouds while flying a plane or a rocket and know from that moment that you're heading off to space. Also, does someone know if anyone worked on turbulence? In earlier versions of the game, rockets would tend to sway and sometimes snap at a certain altitude. Establishing that for the below cloud level and make it especially intense while going through the clouds would make the mission that much more exciting. Aerodynamics haven't yet been established although there are vanilla parts made specifically for that purpose. Parts have wind resistance, is it just used for the aerobraking mechanics? Is this a planned feature in later versions of the game?

Anyway, really nice work. hope to see frequent updates on this

[sarbian]

There's a REASON NWS weathercasting uses some of the most powerful supercomputers in the world for forecast modeling. Even a simplified weather handling system would make the best desktop PC cry.

There is a difference between predicting weather and making one. Doing a coherent weather for one planet would not be that complex if you don't try to do it with a small grid.

[acl8610]

There is a difference between predicting weather and making one. Doing a coherent weather for one planet would not be that complex if you don't try to do it with a small grid.

This is still not a small feat... even generating basic weather so it "looks and feels" like what were used to on Earth is insanely complex, even if only the most primitive equations are implemented. The overall size of the grid dosen't matter, it is the amount of data points within that grid. Even if a "nested" model is used where a coarse global grid is calculated with a floating finer gird within (say around a plane or reentering ship), a personal computer would crash and burn. There is a reason that the weather service uses supercomputers. Where I work, we write very fine-scale weather forecast models this way for small areas less than 20mi and for time frames less than 6hrs. The amount of computing we use is crazy...

EDIT: http://www.mmm.ucar.edu/wrf/users/downloads.html

This may be a good place to start though....

Edited December 3, 2013 by acl8610

[rbray89]

The problem with something like weather is the computation intensive requirements for a 'weather engine'.

Games like Minecraft pull off weather by just applying a global condition. Since nobody can ever see 'the entire world' in games like that, it doesn't matter what the weather is doing from place to place.

But in KSP, you have to realize that applying a 'global event' to the weather, like a world wide storm, gets pretty silly. Thus, you have to break kerbin into areas. Flight Sim X does this and has sectors in which whatever meteorlogical conditions are present, are set for the entire area. While it makes weather doable, the weather is also 'robotic' for a lack of a better term. You fly from one sector into another and it's like hitting a brick wall as winds instantly change direction and speed, and conditions shift from 'decent' to 'horrid' in an eyeblink. And the conditions only affect you when you reach them, so you don't see them ahead of you either. (In real life, you can see rain from a storm ahead of you.)

Ideally, weather would be broken down, not based on the surface, but into moving weather cells. Each cell containing a 'system'. High pressure or low pressure. High pressure systems would be sunny with patches of fluffy clouds and high cirrus, and Low pressure systems would contain various types of inclement weather. Rain, storms and so on.

But even then, you run into the problem of these cells being static in themselves. You see the low pressure area, and it's the same shape and size just moving around the planet. Kind of blah. You have to break these cells down further into sections containing weather patterns such as cold fronts, dry lines, and other meteorlogical disturbances. And at the risk of sounding redundant, you have to break THOSE down into storm cells in order to truly make the weather non-static.

By the time you get down to 'local cell' events, you've stacked at least three to four tiers interconnected patterns on top of each other. Each section of each tier breaking into dozens in the tier below. By the time you get down to the individual local cells of the storm you're flying through, the computer is busy handling hundreds of these local weather cells. And these cells HAVE to relate to each other or you face the brick wall issue and unrealistic patterns. (Winds that blow opposing directions when they shouldn't, rain where rain shouldn't be... that kind of thing.)

Without some miracle of an optimization algorithm, you're looking at a CPU melting weather simulation system ON TOP of a spaceflight physics simulator that's already pushing it. There's a REASON NWS weathercasting uses some of the most powerful supercomputers in the world for forecast modeling. Even a simplified weather handling system would make the best desktop PC cry.

This is still not a small feat... even generating basic weather so it "looks and feels" like what were used to on Earth is insanely complex, even if only the most primitive equations are implemented. The overall size of the grid dosen't matter, it is the amount of data points within that grid. Even if a "nested" model is used where a coarse global grid is calculated with a floating finer gird within (say around a plane or reentering ship), a personal computer would crash and burn. There is a reason that the weather service uses supercomputers. Where I work, we write very fine-scale weather forecast models this way for small areas less than 20mi and for time frames less than 6hrs. The amount of computing we use is crazy...

EDIT: http://www.mmm.ucar.edu/wrf/users/downloads.html

This may be a good place to start though....

My thinking is that if I DO get weather simulation working, it will be with shaders, NOT CPU. This is the kind of thing that shaders excel at. I would use a small render texture with all the channels representing something. Red for temperature, Blue for humidity/precipitation, and green/alpha for pressure likely. Using this, it theoretically shouldn't be too hard to use light from the shader input to "heat up" areas that are lit, "cool" areas that aren't, and generate precipitaion/cloudcover based on input textures (planet heightmap) and humididty. Wind could be read by looking at the green/alpha as a vector. The cool thing is that this should be hardly any CPU if I manage to do it correctly.

[Kilmeister]

Sorry if this has been suggested before.

I'd suggest offering the lights and the clouds mod as separate files. While I absolutely love the clouds,the lights just don't do it for me ,so I had to go and remove all the files pertaining to the lights.

Oh and I can't wait to try the volumetric clouds when I get home. I just read about the update =)

[Hyomoto]

Just reading over your posts, and you are awesome rbray89.

[rbray89]

Ok... So I am REALLY excited. I think I have a few cool things up my sleeve. In order of expected release:

1: Fix for polar UVs. A real fix. None of this pinching/smearing gobbledygook. Thank an idea I got from reading up on tri-planar shaders.

2: Real volumetric clouds. 3DTextures/Shaders. None of this particle-always-face-the-camera gobbledygook. (though I will have to check to see if the KSP version of unity supports this.

3: I thought of how to do real asteroid belts/rings without melting CPUs/GPUs. Imagine rendezvousing and grappling hooking an asteroid with KAS.

[Superfluous J]

Visual Enhancements ®, Now with 50% less gobbledygook!

Looking forward to it, especially snagging an asteroid with KAS. And attaching a docking port to it and a hab module and mining it for ore and kethane?

[rbray89]

Visual Enhancements ®, Now with 50% less gobbledygook!

Looking forward to it, especially snagging an asteroid with KAS. And attaching a docking port to it and a hab module and mining it for ore and kethane?

I was thinking about that... not sure how I feel about it in this mod as it is a "Visual Enhancement" mod... it isn't supposed to be a gameplay altering one. There are a few mods that allow clamping to things, and Kethane/Construction Mods could likely be retro fitted for it.

[Xentoe]

Hmmm sounds awsome.

But how you wanna add asteroids?

(Not even these Asteroids Addition Mod places a belt, you must launch them from Kerbin (blöö how stupid)

But at moment I wait real xurious for the auto spawn for real Clouds this I`m real curious about how it will look :-)

good work till now *thumbie sup*

[skykooler]

3: I thought of how to do real asteroid belts/rings without melting CPUs/GPUs. Imagine rendezvousing and grappling hooking an asteroid with KAS.

I can't wait! Ever since I saw the individual rocks in the rings in Space Engine I've been hoping for this in KSP!

[Umlüx]

could you use your voluminetric clouds to simulate the vapor on the cold tanks you often see on rockets pre launch? i think that would be very cool

[Themorris]

Ok... So I am REALLY excited. I think I have a few cool things up my sleeve. In order of expected release:

1: Fix for polar UVs. A real fix. None of this pinching/smearing gobbledygook. Thank an idea I got from reading up on tri-planar shaders.

2: Real volumetric clouds. 3DTextures/Shaders. None of this particle-always-face-the-camera gobbledygook. (though I will have to check to see if the KSP version of unity supports this.

3: I thought of how to do real asteroid belts/rings without melting CPUs/GPUs. Imagine rendezvousing and grappling hooking an asteroid with KAS.

How do you want to do this? Do you have some pics of first tries?

[rbray89]

could you use your voluminetric clouds to simulate the vapor on the cold tanks you often see on rockets pre launch? i think that would be very cool

You mean the storage tanks? Yeah, I think that would be cool!

[Zander]

You mean the storage tanks? Yeah, I think that would be cool!

OH and while you are working with particles. Can you please do some nice massive fuel explosions when a tank blows up? Would it be possible to make an explosion for every part that has fuel in it and then add smoke trails on every object that survives the explosion? I think this is an easy way to make rocket explosions look really realistic

Heres a titan 34D exploding a few seconds after liftoff