I am confused, what is the logic way of choosing the proper focal length for the 3D viewport? Apart from eyeballing the result that is.
As far as I remember (since forever) I have stuck with 35mm in the Blender 2.7 series and now in Blender 2.8 series with 50mm. It doesn’t bothered me at all.
Is only about the last few years that I started doing sketch practice to develop the artistic eyesight. Is only about the last few days that I discovered this problem with focal lengths and started looking at it. The same model looks entirely different on different focal length.
The problem is deeper rather than just increasing the value.
While on in-animate objects it does not matter so much, for characters, it makes a day-and-night on how you perceive your work piece and the scene, simply put working with the wrong focal depth it will turn out to choose wrong proportions and make your characters look ugly. Producing a character based on the wrong focal length it will turn out to make appear entirely different in each occasion.
So my suggestions are these:
You could do a best-guess of the focal length of the source photography and set the same in Blender. But resist sticking to the default as it is.
The default focal length of the eyesight is 17mm but it looks horrible on the 3D scene. Blender 2.7 would double it to make it 35mm but still not exactly the best case. Also since we always look references from photos and always work on 2D screens default eyesight is irrelevant.
Blender could have some better way to decide on the focal length adaptively. Based on some tests with the default Cube I notice that when you are really far the 50mm is great. But when you close up to the cube as much as possible (to the level of the face) the best focal length would be 100mm.
There is also some very specific theory behind focal lengths, if needed to know.
What is your take on the subject?
I’ve been repeatedly wondering about these things myself, albeit rather in the context of staging, framing, scene-assembly, (image-) composition and layout.
More specifically, I’ve been experiencing difficulties in camera-placement for indoor-scenes, in the sense of ‘how to position my camera inside a room (which is modeled to scale) and still get enough into the frame, without resorting to really weird wide angle lenses?’.
In that sense it has been a bit of a mystery to me how DOPs handle these things in life action productions. Sure, there’s the case of artificially built sets in some soundstage, which might have arbitrary walls removed at will and thus arbitrary extra manovering-space for the camera.
But such a special case aside, I’m e.g. wondering how are car-interor shots done without horrbile wide-angle lens distortions?
Sure the cameras are more often than not mounted on the exterior if there’s a budget of note for things like camera-rigs and sometimes probably mounted to a separate vehicle alltogether (which adds some nice footage-stabilizing overhead, I imagine). Still the obtainable object distances must remain rather limited.
Anyway, maybe (quite likely) I’m just trying to put too much on screen at a time.
It’s not just as simple as focal length, but it’s widely accepted that a ~50mm focal length approximates the human eye.
As for choosing a focal length for characters, especially sculpting faces, you should try and match the reference photos. That said most portraits are made with at least 80mm, and maybe up to 120mm. 100mm is a decent choice.
As for placing a 3d camera in an interior shot, like a living room, I’m not sure why things don’t look right. I have seen it myself and you can never get wide enough and you almost always have to pull back and knock out a wall or something. That’s a really good question. No idea.
That is probably because the human eye has an enormous field of view.
If you want to have the distortion of a 50 mm lens but want the field of view of a human eye you would have to make an exceedingly wide image. This is not practical because for animations you are usually constrained by the 16:9 format and for print stuff you are usually also constrained by the layout.
@Kologe, @Photox, @Lumpengnom
Really good information, thanks for the suggestions.
focal length is indeed important, but so is camera placement.
if you are trying to match your reference image, you need to match both the length and location of the camera. If you need to detail out some lips or the corner of an eye and you move the camera closer in, that is distorting your view, relative to the reference photo.
That’s why increasing the focal length works well when you are up close., it’s functionally cropping the view, relative to the reference image.
Yes, true. The human eye has a field of view of nearly 180°, after all.
One thing to note is you can use an anamorphic lens to cram a wider image/field of view into a given image-width.
The Blender-equivalent of this is the Aspect-setting (which refers to pixel aspect ratio here) in the Dimensions subpanel of the Output tab in the Properties window.
90% of the time I spend actually modelling in Edit mode is in orthographic mode. I hate modelling in perspective mode. Sculpting or organic stuff would be the exception. When I do end up in perspective mode I find the default 50mm to be a good fit.
Only that 2.8 series have an annoying setting on that tends to prevent you from staying in ortho. Turning auto perspective off is important.
I am pretty sure that 2.79 had this auto perspective nonsense as well.
To be honest the only time I find perspective mode useful (beyond its obvious usefulness when modelling has finished or occasional use to check how things look) is when I have the need to view, for example, a closed mesh from inside the mesh to check for weird stuff. Ortho I find more useful most of the time. No battling Blender trying to zoom in far enough with the scrollwheel, no annoyances rotating the view around things. It’s just better.
I like auto perceptive on because since 2.8 it’s possible to automatically snap back to orthographic views with middle mouse button + alt and that almost completely eliminates need in numpad for me. Back in 2.7x auto perspective was a lot more dumb and couldn’t snap back to ortho views so I had it turned off.
Just note what mmb + alt and alt + mmb are 2 different set of similar shortcuts (I disabled second set for this reason).
I never use that alt mmb stuff for navigation. I put front/left/top on “t”, “f” and “d”. That is a lot faster and/or more precise than either numpad or alt+mmb because the fingers of the left hand are allways on these three buttons.
I like doing character art, and 80mm is ideal for portraiture, yet things that aren’t portraits generally look good at 80mm too, so my default is always 80mm. Landscapes can benefit from a narrower lens, but otherwise I stick to 80mm.
Thanks that was good to know.