100% procedural, full scale, earthlike planet

[Heikosch (Heiko Schlein)]

Hello World!

The following raw renderings are the result of my feasibility study to create a fully procedural, earthlike world in Blender.

The biggest challenge for me was to create a procedural cloud system that gives credible results from space as well as within the atmosphere.

The project includes the following features:

• Atmosphere with realistic Mie- and Rayleigh-Scattering and ozone absorption
• Procedural generation of continents including all climate zones
• Credible topography
• Volumetric ocean (continents and islands rise from the water)
• Volumetric, procedural cloud system (in theory also animatable - assuming sufficient computing power and patience XD)

The render times are unfortunately exorbitantly high and are on average between 3-5 hours per image (Cycles at 512-1024 samples, Ryzen 3700x, GTX 1080ti).

Overall, there is certainly a lot of room for improvement in terms of photorealism, level of detail and efficiency. However, I have decided to end the project at this point, since I have already invested a disproportionate amount of time and energy - but in the end I could only make small progress.

I still hope that the current renderings can convince and I’m looking forward to your feedback.

All the best
Heikosch

Edit (April 24. 2023)

As promised, I would like to make my planetary cloud system available to anyone interested.

The principle is the following:

• Application on a sphere with planet diameter
• Scaling of the sphere to the altitude of the clouds (e.g. 2km height)
• Homogeneous displacement of the sphere around the max. cloud height (e.g. 6km cloud thickness)
• Masking of the gained area by cloud structures (cloud domain + isolated clouds)

Feel free to experiment with it or adapt it for your purposes. To achieve even more realistic results, it is also conceivable to use several layers of this system in different altitudes - provided you have enough computing power.

Please do not expect too much. The system is extremely inefficient in terms of render performance and also the node setup itself can be improved in many places.

Small hint:
For best results both from space and inside the atmosphere, I recommend to adjust the volume step rate according to the viewing distance in the material settings (0.01 to 0.001 → lower distance = lower step rate).

You can download the Blenderfile (v3.5) here:
https://1drv.ms/u/s!AiSO9B-8kYCFu94XHYPYDdB3WJ0PsA?e=tVI1GO

@bartv Unfortunately I was not able to upload the file directly here on blenderartists, because every time I tried the following message was displayed:

Sorry, there was an error uploading that file. Please try again.

Am I doing something wrong?

[BlenderMaster15 (Andrew)]

Wow. This is incredible.

[Heikosch (Heiko Schlein)]

Many thanks, Andrew!

[Starfury359]

That is amazing work. fabulous project.

[SterlingRoth]

Very impressive!

[Heikosch (Heiko Schlein)]

Thank you so much!

[Heikosch (Heiko Schlein)]

Many thanks!

[Hadriscus]

Wow, impressive work ! I suppose everything is displaced ? much of the rendertime must be due to dicing meshes and tracing through volumes. What about the terrain shape, did you use procedurals only ?

[Michael_Render (Michael Hickman)]

This is impressive!! How long has this study been?

[Heikosch (Heiko Schlein)]

Thank you for your positive feedback and your interest. Yes, the planet consists of four separate spheres (ocean, land surface, clouds and atmosphere). All modifications including displacement are 100% generated by Blender’s internal shader nodes (no use of image textures and no “manual” influence). The main part of the render times is caused by the cloud system. Since I wanted to achieve good results at all levels of detail, from ~12.700km planet diameter (scale 1:1000) down to 1km line of sight distance, an extremely low volume step rate of 0.001 was unavoidable.

[Hadriscus]

Yes, indeed. Well the clouds are pretty convincing. Perhaps there’s a little room for improvement regarding their shape, but I’ve never been able to make nice puffy procedural clouds myself tbh. The light itself is extremely convincing, how it falls off towards orange, really cool.

Any other issues caused by the immense scale ? I’m working on a set at the moment (an island) that’s 25km², and already at ~5km from the origin my objects start to wobble because of low numerical precision. I had to offset the entire island so that the characters would lie near the world origin.

[Heikosch (Heiko Schlein)]

Many Thanks! From my first steps to the final renderings I invested about 120h (not including render times). A more experienced 3D artist and Blender user with more work routine would certainly be much faster here.

[Heikosch (Heiko Schlein)]

I see it exactly the same way… A real-time preview of the clouds was almost impossible on my hardware under the mentioned conditions. So I had to approach a usable result in small steps. The extreme render times after even the smallest change finally brought me to my knees. Probably my project was a bit over-ambitious.

I can confirm the problems with projects of this scale. The limitation of the decimal places in Blender has also led to unexpected results. In my case a scale of 1:1000 (in the result about 13km planet diameter) was a good copromise and worked well with some exceptions.

[sozap]

Hello !!

Amazing work ! And the time you spend on that seems quite appropriate given the ambition and the results ! Any chances to see some technical breakdowns that would explain a bit more what you did it here ? Or that doesn’t make a lot of sense ?

I always find these experiments quite inspiring !

Congrats !

[bluetygr]

This is very impressive. I have not seen this done in blender before. I agree, those are unusually good clouds for blender’s shader nodes. I second @sozap — would love to hear some more technical details. But yeah, strong work!

[Meshmonkey]

Wowww this is incredible.

[mickmoo]

Love it… nice work… you have great patience…

[Heikosch (Heiko Schlein)]

@sozap @bluetygr

Thanks for all the positive feedback. I will try to create some kind of breakdown over the weekend. I’m not sure yet what makes the most sense here, since the magic basically happens over more or less complex node setups. Maybe I’ll also provide a demo file that comprehensibly illustrates how I implemented the cloud system. I think this is where the most interest is?

For the implementation of an atmosphere there is already a very good topic here on blenderartists.org that I would like to recommend here:

Realistic Atmosphere In Cycles - Artwork / Finished Projects - Blender Artists Community

My solution is not based on a barometric formula. For me the visual, stylistic impression was in the foreground. Therefore my mathemathical understanding is not sufficient XD. However, the basic principle and the physical properties (light scattering & absorption, expansion and density) are quite comparable.

The design of the planet surface was the simplest part for me and is basically a mix of Musgrave Textures (fBM, Ridget Mutifractional) and Noise Textures. If there is interest here, I will try to go into more detail here as well.

[spicy_studios]

This is quite incredible. Wonderful work in all kinds of ways!

I certainly don’t want to ask or push you in any way for this, but maybe just a thought in the name of open source: If you don’t plan to generate any income with this for yourself for now or in the future, and it was just a hobby project, it could be interesting for further explorations to share the blend file to other users. Don’t know if anybody would have interest to develop this further, but maybe you started something that is worth to push further by someone who is interested?

Just a thought, not a request in any kind of way!

And again, I am quite amazed by the results. Very well done!

[piranha4D (Pir)]

Wow.

(Stunning! Those are some fine clouds; I can only dream of making something like that at this point.)