Assets [Released]True Clouds - volumetric clouds for Mobiles & PC

True Clouds is a plugin for rendering volumetric artist-driven clouds, targeted on mobile devices.

Asset Store | Documentation | Android Demo | PC Demo | Official Thread

 

Looks neat, but only you can say how much you should charge since only you know the quality/extent of your asset.

 

Thank you very much. I thought that it won't be so easy)

 

Hello @marserMD

Interesting shader. I have no experience with your prior explosion package, but at least one person found it to be less than optimized based on his comment on the asset page. Which leads to believe the lag shown above isn't from the capture. If you show the scene statistics while capturing - your point about lag from capture can be confirmed.

Since you've learned a lot since initially releasing the explosion pack, are you going to update the explosion pack to be better optimized?

I've used a free capture tool called active presenter and I've found if I increase the capture frames to 60 fps it seems to perform better.

 

Hi, @theANMATOR2b

I am terribly sorry for such a huge delay. I have worked hard on the shader to make it work on mobiles in full-3d, which was hard since mobile graphics commonly support only ARGB32, some of them don't support depthTextures and so on. But I made it through. And ofcourse I worked on it's quality.

I am not going to update TrueExplosions, but I might want to create a new explosion tool later. With better effects AND better optimisation.

Here is a video to show that my plugin works really fast on pc. On monday I will publish photos of android profiler. I don't have the right cable at home to connect my phone) without framecapture and with max settings it takes 0.7 ms to render clouds on my laptop(old core i3 and nvidia gt530 m)

 

Hey, guys. I've sent the package for approval!
Plugin's quality has majorly increased in past month. Here are a few pictures that I got easily out of my plugin:

 

Two years later there still is no volumetric clouds asset optimized for mobiles on the Asset Store!
That's why I decided to finish my job. Previously my package was declined (most probably because of poor UX), but now I'm much more experienced and ready to bring it on
Update:
1) Better UX/UI
2) Performance optimizations
3) A way to render noisy clouds! They look freaking awesome and have only a small impact on performance (compared to simple clouds)!

I don't have time now to make a video, but it looks even better in dynamics.
What do you think guys, does it look like a cloud to you?

P.S. ETA on the asset store: 1 month + approve time.

 

Today I've fixed issues with deferred shading.
One step closer to publishing

Next on the schedule: adding an option to test depth less accurate, but faster to improve performance on low-end devices.

 

Figured out a way to drastically optimize depth testing without noticeable look change. I'm getting close to 6 ms for clouds rendering on my nexus 5x!
Fixed bugs related to noise reprojection.
Next on the schedule: better noise for density and better profiling for mobile devices. Unfortunately, the only way to correctly measure the impact of clouds rendering is to measure the difference between enabled and disabled rendering fps.

 

this is so cool! do you need beta testing? I will try it out on ipad mini 3, iphone 7 plus, apple tv 4 and apple tv 4k.

 

Yes, I do, good sir
I'm a bit lazy in setting up apple build environment, so if you would assist me in a few simple steps: build, check that everything is OK, send me performance statistics(will be shown on-screen), then you will receive a free copy of TrueClouds even before release and a promo code for free copy after it is released.

You might need to do this several times though because some bugs might occur But I won't abuse your help, if something goes terribly wrong, I won't use you as a build server

So if you are interested, just add me on skype (same nickname as here).

 

The sweetest cherry of any image effect is the support for edit-time preview.
Done

 

Nice work!

 

What does every GPU intensive asset need? An easy-to-understand built-in profiler so you could easily understand what impact on performance the asset makes. And why would it be hard?

1. Clock rate might change over time, especially on mobile platforms
2. Framerate is capped at 60 fps, so it is unclear what part of "56 fps" took clouds and what part was taken by, for example, skybox.
   
3. Some GPU tasks may or may not be executed asynchronously on different platforms.

But what matters to every one of us is the difference that the script makes.

So here is what I came up with:

1. Wait for some time to pass since app start-up.
   
2. Perform test.
   
   • Make 2 * N batches, each T frames long.
   • In N randomly selected of those, enable clouds rendering and measure performance.
   • In other N batches, disable clouds rendering and measure performance.
   • To address the 60 fps cap, add workload of 16 ms for each frame during testing.
3. After the performance info is collected, print out median and 90 percentile of deltas between batches with clouds enabled and batches with clouds disabled.

For my Nexus 6x it takes about 6.4-6.7 ms to render the clouds with all features enabled.
Now I can easily judge performance impact that certain features make.
And I will publish a demo app this week so you can decide for yourself whether you'd like to use my asset

 

What is crucial to an asset that makes heavy use of noise textures?
I believe that a convenient way to generate appropriate noise textures is a must-have because it is the simplest way to change the appearance.
So these are some pretty images I got out of my generator

And yes, they look like abstractionist's watercolor work, this makes a lot of sense when you take into account that each color represents a point in a 3d sphere with radius 0.5 and a center at (0.5, 0.5, 0.5)

 

http://momentsingraphics.de/?p=127

Som interesting article about noise

 

Thanks! Looks interesting. I might apply some of the ideas.

 

More cool stuff with the same kind of texture

http://ellenack.com/2016/11/05/grass-geometry-shader/

 

Late night teaser.
Created this scene in about a minute just by using a mountain from the Asset Store, a bunch of spheres and "True Clouds" plugin of course.

And remember: it is fully mobile ready. These are the quality settings that your mobile device can handle!

Development update:
I've fixed all known bugs, improved the interface as far as possible, cleaned up the project.
Most importantly I've fixed triplanar projection issues, which led to a huge quality boost.
ETA for submitting package -- this weekend.

 

Fixed another noise projection issue. Those just keep coming!
And the quality improvement is insane.

 

Well, it was a rough night
Creating an asset is only half a way if you are a programmer. For me showing the asset off is the hardest part.
I know how to write code and how to fix bugs, but I don't know how to judge my own "art"

But when the street lights turned down and the night began falling back, I have submitted the package to review.

My footage folder has eaten up 20 GB of HDD and ~20 hours of my life. The funny thing is that the clouds got set up in minutes. But composing the footage together was a living hell

 

Looks awesome!

 

After a month of waiting, "True Clouds" was finally accepted!

 

https://www.assetstore.unity3d.com/en/#!/content/107379

 

Russian speaking fellas, I've made a post describing the algorithm behind the "True Clouds" on Pikabu.

Dear Reddit users, which subthread should I use to make a similar post? Can I add a link to the asset? Should I go in-depth with every tiny detail, or keep it more entertaining in an overview manner?

 

quick question, I need clouds in my F-22 mobile game, very much like your example, but as a high altitude "cloud blanket", extending to the fog fall off, flying above and below, with camera able to pitch, yaw and roll..while the clouds maintain the horizon plane.

And looking through a mobile transparent shader (cockpit glass & HUD)

Show me your asset doing this and...SOLD!

(maby using unitys example air craft, just curious)

p-

 

Can you please clear out one thing: is the camera going to fly through the clouds or not? Because "True Clouds" don't support flying through

Looking through a transparent glass/HUD is a non-issue.

As for creating a demo -- I might consider adding cloud blanket to an existing demo since it's just a few minutes of setting up and placing the clouds. But creating one from scratch is a bit too demanding for an individual sale

 

so, you cant put this on a standard first person camera/controller and walk through the above scene?
if you can walk through it, you can fly through it...

 

The key point here is that a camera INSIDE a cloud would render invalid images. A camera outside of a cloud, on the other hand, would render a correct image whether you are walking or flying.

This might change in further versions, but currently rendering from inside a cloud is not supported.

 

would be cool to be able to go inside, might be good for a future version indeed

 

I'm afraid that correctly dealing with it while maintaining the performance is nearly impossible. So there must be some clever hack done, which I haven't come up with yet. I might come up with one though

 

What would such an 'invalid image' look like?

 

That depends on the mesh.
If it is convex, something like this would appear:

If it is not convex or there are multiple intersecting meshes, something like this:

The point is: rendering from inside the cloud is not supported and might not be supported at all.

Anyhow, you can track the progress on my trello board https://trello.com/b/DZ4wTayS/true-clouds

 

Fair enough. I was wondering whether it was rendering something at all. Looks like the object needs to be disabled, though. I'm wondering whether simply using a screen space fog effect if the camera is inside or close to the cloud mesh should work. As long as the mesh is convex and can be reasonably approximated with a cube or a capsule, that is.

 

That is pretty possible -- you could disable clouds rendering on entering cloud's trigger and enable a dense fog. Although I don't think that such functionality should be a part of "True Clouds".

 

Looking for VR testers https://forum.unity.com/threads/givaway-true-clouds-volumetric-clouds-for-mobiles-pc-vr.513776/

 

VR testers found

 

Have released an article on a Russian IT resource about rendering technique of my volumetric clouds.
Got in top trending in no time!

I'm going to make an article on Reddit as well, but don't know what is the best way to create an illustrated tutorial.

 

Good news:

• My teaser on Reddit got a bit viral!
• An update is coming up featuring depth filtering for blur!

Bad news:

• I got scammed into giving away a voucher. And I don't want to give him any updates, so we'll have to wait for his promo code to be blocked

 

this looks super cool, reminds me a bit of the tech we had on watch_dogs 2 for the fog in the SF bay area. Nice work

 

Do you mean that you have worked on WD2? If so, It's an honor to have you in my thread

 

I have made an article about how TrueClouds work!
https://medium.com/@marsermd/true-clouds-how-is-it-done-b15f6d9a5167

 

True Clouds v1.1 featuring Distance Blur Filtering will be out soon!

 

Can I disable rendering of just one cloud when I fly in it?

 

Are there any pics or videos showing the UI or process to set up? I'm curious to see what the controls and options are.

 

I am a bit disappointed by this. It looks fantastic in the demo scene, and is very well optimized for that specific use case, but seems to have some serious limitations outside of that which prevent me from really using it. If I am using this wrong, please correct me.

1. At the screen edges, there is a very unnatural fading in/out of the clouds. For my purposes, the camera needs to be able to pan around the scene and fly around the clouds (doesn't have to fly through), and this looks very odd.

Here is a video of the problem with the screen edges

2. Blur is extremely dependent on size in screen space. Small clouds, such as those you would use in particles, simply disappear after a short distance with blur settings that look good at close. Meanwhile, to get them to look good at a distance, it requires a very low blur setting. . I think this is because they are getting blurred out.

Here is a video (with 50m wide cube for reference, draw distance at 5000m) where I start far away with a small blur distance. It looks good until I get close. I adjust the blur so that it works well up close, but then at a distance it no longer works. Worth mentioning is that I had to create a new instance of Cloud Camera 3D, and a totally separate layer, to support both the cloud wall and these smaller clouds. There are simply no settings that work well for both.

This really is my biggest issue, but the distance filtering you posted earlier might solve this issue.

3. Not very robust in general for large distances or with having vastly different depths. For example having kilometers wide clouds in the distance and up close is just difficult to find settings that look good for both situations. This relates to the above problem, but it ties into noise as well. They tend to have the same level detail despite the large difference in distances.

Compare the wall on the left, which is at 10k, to the one on the right, which is at 30km.

Those are my more addressable problems. The rest of my concerns I would be okay with the OP simply telling me are out of scope for this kind of asset. After all, it does seem to be highly specialized for static rendering on mobile, which it does excellently. However, it might not be quite what I'm looking for.

My pie in the sky wishlist (And a lot of why I'd love to learn how to do this myself.) is as follows and would make this worth every penny for myself. Again, I am completely okay with being told that all of this is out of scope, so feel free to ignore it if you like.

1. More robust support for large distances. (Blur + noise being based more in world space than screen space)
2. Ability to configure clouds on a per object basis rather than for the entire screen. (Or perhaps by material?)
3. Hand in hand with the above: support for particle coloring/alpha so that this can be used with particle systems. I'd love to be able to use this shader on meshes generated by particle systems.
4. More precise depth detection. When using a camera with a large view distance, blending by depth looks extremely bad due to the loss of precision. I understand this is an issue with storing depth in the alpha channel, but maybe there's a better way. (Would a logarithmic depth buffer apply here in some way?)
5. Ability to take light from arbitrary and non-directional dynamic light sources. (Not that important to be honest, but it would be nice.)

 

If you disable the cloud's renderer or game object, It would stop it from being rendered. You can detect entering a cloud by using physics triggers

 

Words of gratitude
Thank you very much for posting here your problems precisely and not just raging out that the clouds don't work. Your feedback is extremely important to me as this is the only way to know what my users struggle with.

To your concerns

I have to confess that the sky in my first trailer is done with a bit different shader. Unfortunately, It turned out that I can't ship it to all mobile devices, so I had to remove it.
I have come up with a different technique though. Added in my trello. Please do tell me if what is shown in the trailer matches your needs and I can give you a dev version if you can't iterate without this feature. Just send me an e-mail at mischapanin@gmail.com with your invoice number.

Depth blur filtering should solve this issue, although it produces some other artifacts and requires picking the right Blur Power and Blur Threshold. Please do post here whether it fits your needs or not.

Noise size is actually scaled as well, but without the depth blur filtering it might appear that it is constant.

To the less important ones:

This is what you've mentioned previously, right?

This is close to impossible since this is a screen space technique. Although you could have two Cloud3d components with different setups for different layers (see Dragon demo). This is a limiting approach but works well for some scenarios.

You should be able to do it right now (without different settings per particle, see previous paragraph)

I actually encode depth in R and G channels. I could at least increase precision by 256 times (as an option of course since it will slow down the whole effect a bit). Added to trello.

This is on my list already as I think that it will allow for more general-purpose usage of True Clouds, giving more artistic control and thus hugely increasing its value. Most probably it will be implemented similarly to a deferred rendering technique.

As a conclusion
Thank you for buying my asset. I can't believe that it has such a huge success. My asset actually shows up first when typing clouds!

All the tasks marked with feature/bug on my trello will be completed. True Clouds has such a low price because I realized that I might have failed to grasp my customer's needs. After I feel that the customers are mostly satisfied, True Clouds will cost about $35. So thank you for supporting me at the very start of True Clouds on the market.

P.S. in your video the cube has vertical lines on the edges. To fix that you might want to increase Blur/Threshold and Blur/Power.

 

Hi. I have planned to release a tutorial but it seems that the interface and functionality still might change a bit, and making tutorials is extremely hard So an official tutorial will be released later (don't have any ETA yet).

Here are screenshots of the editor interface:

And the noise generator:

Some parts of that are dynamic, but it should give the impression along with the docs.

 

BTW, where did you find out about true clouds?