Are there any disadvantages or problems when rendering/lighting in a non-real-world scale?

I’ve always made my models about 10x of real world measurements. A head I made fits roughly inside the default cube (2x2m, while a human head is ~20cm).

Are there any disadvantages to doing this? Do lights work differently at this scale? Should I start working in real world scale? The only things I know about are:

  • subsurface scattering is scale dependent, but I don’t use it often enough to see the difference
  • depth of field f-stop needs to be drastically adjusted to see even a bit of blur
  • Eevee contact shadows have a limit defined in meters, so there can be weird artifacts when out of scale

Anything else?


I always try to scale my scene to more or less correct dimensions, because attenuation of direct light, bounced light, reflections, refractions and more are all based on real-life formulas in realistic renderers. I guess correct scene dimensions matter less in EEVEE, but you’ve still got the correct light attenuation (inverse square falloff I believe).

Light behavior in realistic scene dimensions also matters for things like glass. If your glass has 20 cm thickness, light will not penetrate through it realistically, resulting in glass that is too dark.

Also, never forget to press Control + A and apply the scale after scaling, otherwise the renderer will still regard the scene elements in the wrong scale.


You can scale everything and there is scale property how many meters one blender unit is. It doesn’t matter in light. Physics are more sensitive to scale in blender units, not in meters.

1 Like

Scale doesn’t affect modeling but it’s important for everything else, specially for VFX when goal is to match real world reference. For example, huge assets have a insane amount of UDIM’s to get enough detail to looks like it’s big, so scale affects texturing process. Proper lighting will remove that toy look when objects are too small, and will get that depth for very large models. By working with correct scale you will only get advantages. Correct depth of field, motion blur, simulations, it’s easier to get a nice looking render and interchangeability between softwares.


It does matter to light, you will have to use much higher intensities for local lights if your scene is much bigger.


A bigger scale also requires longer rays for ray casting.


Blender base unit scale should solve that?

Well, yes, the Power values will adjust to the scale of your scene. But then you can’t rely on physically based values anymore. You’ll have to guess the values you are entering (or do some math to divide them by the square of the unit scale).


I think bump is also scale sensitive.
But I suspect that all of this is kind of relative and also bound to how precise you need to be.

SSS will change according to scale, but do you know proper value for radius to enter ?
Sometimes you also need to cheat the scale, like say you’re working on a watch or the interior of a smartphone, probably you want to change scale so 1 BU = 10mm or something like that.

Same with huge exterior, if you plan to do a city you might change the scale so 1BU = 100m.

Even if some parameters will obviously need to be adjusted to that new scale I’m sure all this is pretty relative, and once you have correct value for DOF, lights, bump etc everything will stay consistent.

If you have real life things to match having correct scale is pretty useful.

Given that your work is quite stylized I’m not sure you’ll have real benefit on using real life scale and it’s ok to eyeball things especially because you aren’t looking to reproduce reality but you’re making a stylized version of it.

That’s my two cents but I didn’t really look into that problem a lot.


You must multiply your Scale and Distance values. Example, if an object is 20 cm in real world, but you create 2 m., then you must multiply by 10.

And, for Displacement, if you Applied your scale, you must do same thing, but if you Scaled but not Applied, you must do not…


There is old rule that is best to model in real scale. Or at least that scale model to real scale latter when ‘arguing’ with polygons is over :smiley:
Many reasons, people already mention lot of them. But again, as far I know all renderer love real world scale. Simply they are made in this way.


Ok, looks like I’m modelling to scale from now on :stuck_out_tongue: Thank you for all the answers. I’ve been doing some googling on this topic and there are posts from literally decades ago saying “model in scale, stupid” I guess I just needed someone to say it directly.

I’ve been sticking to the 10x scale because over time I’ve made for myself a bunch of common-use shaders, base meshes and armatures, and they almost all break after I rescale and apply with ctrl+A. That’s probably because they’re quickly built hacks that just barely work for the specific task they were made for. I think programmers call this “technical debt”, and it’s time to pay up ;(


My point is that assets can be created to specific scale and that is actually better than trying to world in 1m = 1bu scale

Physics are scale sensitive on blender units and when that is wrong, it is very clearly visible.

My opinion is that It is better to split scene to layers and choose what is ideal scale to each layer. It does matter are you modelling ants, humans, trucks, mountains, planets…

3D modelling is very close to miniature or movie set building. Choosing scale is important.

1 Like

This is a slight tangent, but still related to the original question…

Physics simulations. I’ve not done a lot with Blender’s dynamics, but have used Bullet in Lightwave before. I recall that was programmed to work best with objects in the region of 1m, so when doing simulations with objects that were only a few cm in size needed to be rescaled. Once the simulation was done, the results were baked, and everything could be rescaled back to real-world scale if it had to be rendered with other items.

What about Blender’s rigid body simulations? Do they work with tiny objects or does it also work best with 1m-ish sized models?

I think that makes sense and that’s what we probably naturally do when working on character only.
The default cube is 2 meters high and seems a bit small in the 3D space when just working on a character.

But it’s possible to adapt to that scale, probably by tweaking the grid, camera clipping…

At least , having a fixed scale like 10 BU = 1 meter is always good, so you can always convert things easily from one project to another, rather than having a complete arbitrary scale.

I you want to play with physics yeah maybe that’s something important to take care of, but scale to me is always project dependent especially when they tend to be different from one to another.

I had a really hard time on one project where I needed to have the sun, earth and a satellite in the same .blend scene. Client wanted everything on scale which was impossible from the start. Sun was 2 times smaller (and closer) and the satellite ended up being 10km long. But even with these I had too many rendering issues, some simple stuff was really impossible to make.
I could have made my life much more simpler by cheating thing way much more there, that’s the lesson learned from that project.


Ideal is that smallest objects are 0,2 BU and lähestyy 5 BU, and some static collision mesh need to keep largest face below 10 BU.

Very small objects and very big difference in objects sizes causes issues.

1 Like