It deserved to be revived because its the best clouds I have seen in blender to date, and they are animated! Then I made a bounding volume that has the shader. If you turn off the physics for the particles then you can apply modifiers to distort it. It just kind of defeats the purpose of what I was going for specifically, which is a completely shader based cloudscape with easy to use parameter inputs.
If you want full control and spend more time to get the right shape, particles are probably the way to go. I agree purely procedural is preferable if only for ease of setup. I want to revive this topic once more now that I finally played around with the shader again. I have to admit that my interest in this 1y old project got triggered by that new real sky add-on that just came out on the blender market.
Honestly I made no changes to the shader itself yet. I threw it into 2. It works really well with real-time. Here is an animation I rendered in HD with a render time of only 4s per frame:.
You have to know that this sequence would have most likely not even rendered in cycles my GTX crashed when i tried. With Eevee I can really imagine picking this up again and making this into a polished shader with decent documentation. Let me know if you are interested. I am interested. This looks great.
Easy to set up setup. Single clouds for hand placing. Generation of whole animated cloudscapes. Paintable cloud distribution CloudsWriting. That might not happen before a later release though. I hope you enjoy it, definitely show me what you create with it! You can tweet me simonthommes or just post it here. This is fantastic. Thank you very much. Im getting there I think I am. THese are either getting much better or im going blind.
Spring — Blender Cloud
Just trying the last ones in a landscape. Ill move on to other cloud layers once i tighten this a bit further. Can you please give us some insight on how you achieved these non volumetric procedural clouds i would like to create a ocean scene like this with a deg sky and clouds for animation thanks photo 1. If anyone was interested, I just made a tutorial showing how to create fairly realistic procedural clouds in Blender non-volumetric. Instead of doing everything in the world node group, I applied the cloud material to a large plane instead.
Non Volumetric procedural clouds Artwork. Blender Tests. Bunc Bunc May 1, , pm 1.
- LBJ and the polls.
- Multiscale real-time procedural animation of clouds and cloud systems?
- The Greatest Story Ever Sold: The Decline and Fall of Truth from 9/11 to Katrina.
- Related Posts.
Here is a typical example of what I can produce so far. Bunc Bunc May 1, , pm 2.
- Principles and Parameters in a VSO Language: A Case Study in Welsh (Oxford Studies in Comparative Syntax);
- Dynamic Volumetric Cloud Rendering for Games on Multi-Core Platforms?
- The Clockmaker: The Sayings and Doings of Samuel Slick of Slickville (New Canadian Library (Paperback)).
- Fundamentals of Radiation Materials Science: Metals and Alloys;
Bunc Bunc May 1, , am 3. Bunc Bunc May 1, , pm 4. Bunc Bunc May 1, , pm 5.
Sketch Based Volumetric Clouds
This texture is filled by integrating each row of each column and dividing the values by the total value of that row. But before dividing each value, we will also store each total value in the 1D marginal function texture. For brevity this textures are named after pbrt nomenclature. The 1D marginal textures gives us the cumulative distribution and the avarage values in y-axis avarage of rows in a column.
Hence we use a seperate 1D texture to break this dependency cyle. If we choose a random value of 0. If the cdf was linear the value would be around This respectively shifts our probability to choose a column in lower numbers. We can visualize this by taking a look in a violin plot that shows the densities of samples choosen based on this cdf. You can clearly see that around we have much more samples because the avarage density based on rows are much higher in that region. Now that we have a column number we can go ahead and choose a row based on the 2D conditional CDF.
Finding a row follows the same procedure and gives us the row number that is most likely to be choosen. After we found our column and row numbers we can visualize the points on the image. As you can clearly see sampling is much denser in the regions of luminous areas. And fewer at the lower hemisphere where there is not much light to sample. Finding a ray direction from here on is straightforward.
This gives us the azimuth and elevation to convert them from a spherical coordinate to cartesian coordinate. To find the pdf values we first divide the 1D marginal function value at choosen column and divide this by the total value of said texture.
Then we find the conditional pdf by value at 2D conditional function texture at choosen row and column, and divide it by 1D marginal function value at choosen row remember these were the total values for rows. Then we simply multiply marginal and conditional pdf values.
Finally we can visualize the directions and pdf values in a nice plot. The following 3D graph shows the directions on a unit sphere, colored according to their pdfs. Remember that the higher the pdf the lower the light contribution.
We can clearly see the outline of the sun and the area below it is more dense in samples.