PDA

View Full Version : Camera, lights and ...model scale and render resolution?



Aww167
12-13-2016, 04:19 AM
I have a question that's been floating around my mind for a while and have not so far found a definitive answer in any of the reference material or online discussions I've come across, though it's not unlikely that I might have misunderstood or overlooked it, but...hopefully a few more knowledgeable and experienced people on this forum could help cast some light.
Is there any direct correlation between the scale of a model and the ideal resolution of the completed render, or is it all just relative and a case of finding which camera settings work best for the particular scene at hand? As a relative newcomer to 3D and LW I'm still grappling with the render technology though am gradually absorbing more and more useful info and experience. I've been using real-world measurements for the objects I model but this is mostly because it helps achieve correct proportions more easily - I feel sure many more experienced users can judge by eye whether something is correct, but I wonder whether there's any intrinsic value to keeping things at real-world scale, especially when it comes to render times and also when seeking to obtain a pleasing perspective on a given scene? Would anyone here consider that varying the model scale produces better or worse results in any given scenario?
My instinctive feeling says it's really about matching up the right camera settings and image resolution to obtain the desired result, but I can't help wondering whether model scale is also a factor?

daforum
12-13-2016, 05:18 AM
One way to scale when modelling is to find a cg model of an adult figure ('ideal adult' I think they are called) and work around and to the figure scale.

djwaterman
12-13-2016, 05:25 AM
Keep modeling to scale, all the lights and cameras are designed for things being at correct scale. If your doing something mammoth, like a space scene with fifty mile long space ships and huge planets, then pretend you're making miniatures for a movie and scale down these objects to more manageable scales just for convenience, big enough to photograph and light realistic but small enough so that you''re not having to navigate in kilometric distances. Like-wise if rendering molecules, it would be silly to try and model these to scale.

Aww167
12-13-2016, 07:41 AM
Keep modeling to scale, all the lights and cameras are designed for things being at correct scale. If your doing something mammoth, like a space scene with fifty mile long space ships and huge planets, then pretend you're making miniatures for a movie and scale down these objects to more manageable scales just for convenience, big enough to photograph and light realistic but small enough so that you''re not having to navigate in kilometric distances. Like-wise if rendering molecules, it would be silly to try and model these to scale.

Thanks for the reply. I've been mostly attempting archviz and have downloaded a few models to examine for reference, almost all of which are many times larger than the real-world equivalents so I wondered if there was a benefit to doing so, although I can well imagine an experienced modeler probably isn't going to need to resort to measuring all the time. I'm still at a fairly basic level of understanding how to get the desired render so am experimenting a great deal but thought it would be good to have a more experienced view on the subject.
Your comment about modeling micro-organisms does make the point about scale rather succinctly though!! I've read that certain renderers work better on real-world scale but not come across anything regarding LW in that respect.
Thanks again.

MonroePoteet
12-13-2016, 08:14 AM
I always try to model at real-world scale, except for the really big and really small stuff referred to by djwaterman. This doesn't really have to do with rendering, but more with scene composition and organization. Trying to put a real-life sized space station (for example) around a full sized planet is a real pain. For example, the diameter of the earth is about 6000 miles, while the ISS orbits at around 249 miles, which means in terms of the relative scene size, it's almost skimming the surface. Trying to put a real-life sized building / city into a real-life sized landscape is also troublesome. In both cases, I spend a lot of time zooming in and in and in and in or out and out and out using grid size changes between the human-sized objects in the scene and the real-life environmental objects.

The LW camera doesn't have a real aperature which would affect depth-of-field of miniatures or human-sized scenes - you set up DOF artificially and can use settings which aren't possible with a real-world camera. As well, you can set focal length or horizontal / vertical Field Of View with no relation to what a real-world camera might be restricted to.

One real-world reason for not modeling very small stuff to scale is the inability of the 32-bit floating point representation of numbers to store continuous values with small changes in the 4th or 5th digit of accuracy. So, if you model something where there are important features that require numeric differences at that small scale (e.g. 0.000457 is important to differ from 0.00046), then LW (or any other software using 32-bit floating point numbers) won't be able to store those small differences accurately.

RE: render times, there shouldn't be any difference if the scale is accurate or not. The numeric calculations performed by Layout will be exactly the same, just with different numbers.

RE: importing objects that are huge, different packages use different default scales. I usually import them (e.g. DXF, OBJ, 3DS), and then scale them to real-world scale in Modeler, save them as an LWO before loading them into Layout.

Just my approach, and I'm just a hobbyist.

mTp

djwaterman
12-13-2016, 08:24 AM
Most of your downloaded models from other software will come in huge. What I normally do, if I know the basic dimensions of the object, I create a box in another layer, type in the width, height and depth values and then with that box being in a background layer as reference, I grab all the over-sized object and size it down to fit inside the reference box, then delete the reference box.

gerry_g
12-13-2016, 09:21 AM
all assets stored at real world scale saves a lot of hassle, have a low polly mannequin 2 meters tall that have loaded in a layer when I model, amazing how much that helps with proportion

Aww167
12-13-2016, 10:12 AM
The LW camera doesn't have a real aperature which would affect depth-of-field of miniatures or human-sized scenes - you set up DOF artificially and can use settings which aren't possible with a real-world camera. As well, you can set focal length or horizontal / vertical Field Of View with no relation to what a real-world camera might be restricted to.


Thanks for reminding me. That's something I need to explore far more. One of the issues I had in mind concerns achieving a desirable perspective in an interior and my natural preference is more for seeing through the eye than through a lens which I find restrictive. Actually, I sometimes think even that's too restrictive! Focal length and zoom factor in LW camera obviously aren't the result of seeing through glass, but I assume that approach is the best way to resolve 3d space to a 2d image. I'm still unclear exactly how far the analogy is necessary and am no doubt overlooking many more creative ways to achieve results.

Aww167
12-13-2016, 10:35 AM
Most of your downloaded models from other software will come in huge. What I normally do, if I know the basic dimensions of the object, I create a box in another layer, type in the width, height and depth values and then with that box being in a background layer as reference, I grab all the over-sized object and size it down to fit inside the reference box, then delete the reference box.

That does make sense and I feel instinctively it's important to establish good consistent work practice, especially with something this complex. I don't want to make choices that might restrict the quality of the outcome, then again, the stage I'm currently at makes it all seem overly haphazard at times , but then experimentation is at least half the fun of using this technology anyway. I don't want to get stuck in a rut by avoiding trial and error, but I do also like to get the results I desire without too much time and effort. Choices, problems, choices, etc....:)

prometheus
12-13-2016, 11:33 AM
Keep modeling to scale, all the lights and cameras are designed for things being at correct scale. If your doing something mammoth, like a space scene with fifty mile long space ships and huge planets, then pretend you're making miniatures for a movie and scale down these objects to more manageable scales just for convenience, big enough to photograph and light realistic but small enough so that you''re not having to navigate in kilometric distances. Like-wise if rendering molecules, it would be silly to try and model these to scale.

And if we try to bake hypervoxels clouds, you would need to rescale down all models ..since baking hypervoxels 1meter in size may not work well to render, and with rescaling you would probably also get a wrong lightshadow softness angle from the sk sun, which may need to be scaled down as well.

Amerelium
12-16-2016, 05:15 AM
I use LW to create scenes based on Iain M Banks culture universe, which means dimension ranging from room size to hundreds of kilometers, millions in case of orbitals. LWs scaling options on of the reasons I went for it way back in the day.
My experience is that it is always better to use real scaling, just to be able to better keep track of complex models and scenes.

More importantly, in LW there is no reason NOT to...

prometheus
12-16-2016, 11:49 AM
I use LW to create scenes based on Iain M Banks culture universe, which means dimension ranging from room size to hundreds of kilometers, millions in case of orbitals. LWs scaling options on of the reasons I went for it way back in the day.
My experience is that it is always better to use real scaling, just to be able to better keep track of complex models and scenes.

More importantly, in LW there is no reason NOT to...

As a guideline yes, but you canīt really use real scale dimension if your in a situation I described the post before yours about baking voxels, it thing there may be obstacles as well regarding extremly large scales involving enourmous suns or galaxies.

bullet similations may also work better at a different scale depending on the item for bullet dynamics to work best or most realistic, which may suggest you may need to scale down and up after the simulation to the items real scale, if it is extremly small or extremly larg from starters.

Amerelium
12-19-2016, 02:50 AM
As a guideline yes, but you canīt really use real scale dimension if your in a situation I described the post before yours about baking voxels, it thing there may be obstacles as well regarding extremly large scales involving enourmous suns or galaxies.

bullet similations may also work better at a different scale depending on the item for bullet dynamics to work best or most realistic, which may suggest you may need to scale down and up after the simulation to the items real scale, if it is extremly small or extremly larg from starters.

I saw that, so I ditched the whole baking idea.
What I learned is that if I keep my distance from the clouds, the render time is minimal - like 10%. Can get quite good results by basically massing about with just the amount of solid fill points and particle size values. Once I get within a certain distance at during render the AA passes begin to take a LOT longer. But by keeping the distance it's almost like there's no hypervoxels to calculate, even with ships flying through them and stuff like that.

Perfect for my needs.

In my latest project I do everything at a 1/100 scale though, simply because modeler cannot zoom out enough to make working on a 10M km diameter orbital practical.

bobakabob
12-19-2016, 03:54 AM
You can get away with different scales in Lightwave, but best to keep consistent for workflow as everyone has said. For Archviz, you can build up a uniform model library and so create the illusion of cities very quickly. For Animation, RHiggit (and pretty sure Genoma) is designed to be used to scale so it helps if scenery models match to start with. You really do get some wild deviations importing from different sources...

djwaterman
12-19-2016, 04:43 AM
Matt has a script on his site that allows you to scale an item base on two selected points, so if you know or can guess the dimension of some part of the object, and select it's two end point and type in the value, the entire object will scale down or up accordingly. It's called set scale. Could help when resizing objects from other 3D apps.

http://www.pixsim.co.uk/

Aww167
12-19-2016, 12:57 PM
Matt has a script on his site that allows you to scale an item base on two selected points, so if you know or can guess the dimension of some part of the object, and select it's two end point and type in the value, the entire object will scale down or up accordingly. It's called set scale. Could help when resizing objects from other 3D apps.

http://www.pixsim.co.uk/

Very helpful to know about that and a handy solution for scaling matters. Thanks very much.