Understanding Focal Length of a Portrait

killswitch · Dec 8, 2021

Hi,

I am a 3D artist working in the game industry, and we often get to recreate or cleanup scans of likeness of actors and actresses. Most cases we collect reference photos from various angles to understand the structure and form of the head. For full frontal shot of the portrait, we try to decipher what focal length was used to take the picture. This helps us to set our camera in our 3D software properly, otherwise the 3D likeness of the portrait looks fine from front, but breaks apart when we turn our 3D model and looks completely off. This is part our job to sort of guestimate and line up things as closely as possible.

My question is from a journalistic point of view, and the distance allowed to photograph a person at typical red-carpet location, or people on small stage posing for let's a movie premiere

What's the most commonly used focal length used for headshots and half body length.
What is a good process of deciphering what focal length was used from just reading a portrait photo. Like, typically how far the ears get tucked behind a face due to distortion when something like 24mm-35 is used vs how flat it can get when a 70-85 mm used in the above mentioned scenario. I know this is a very open ended question, I guess what I am trying to get at is, what are the clues I should be looking at. Like the most tricky part is the width of the jaw is hard to evaluate from low-res photos from internet.

Any guidance is appreciated. I remember seeing an image where a photographer took portrait of a woman using 10-15 different focal lengths to illustrate how much distortion and flatness we can achieve using lenses from 24mm to 200mm.

Sporgon · Dec 9, 2021

My observations on this:
2: from just a portrait picture without any further reference I don’t believe you can tell the focal length that has been used. Certainly from a perspective point of view there is no real difference from 100mm up on FF. But then I’d challenge anyone to try and tell 70 & 100 apart. In a tight head portrait 50 begins to exaggerate near and far, that is bigger nose, smaller ears. Beyond 35 can become extreme but then it can be deceptive; I’ve seen 28mm head portraits that I wouldn’t have though were shot on that focal length. You’ll see a big difference between a head portrait at 28mm and the same at 85mm - assuming you had the two side by side

1: if you are wanting to flatter the subject in a tight head portrait most would want to use 100mm or greater. For a half length portrait 50 would probably be most common, but it depends on how much space was available. I wouldn’t like to guess between 35 or 50. Longer focal length will help isolate subject for a given aperture.
In short I think you’ll struggle to do what you’re hoping.

YuengLinger · Dec 9, 2021

For the level of precision and automation used to create believable, higher-end video games, I would guess there are programs that can analyze a series of images taken from wide-angle to super-telephoto.

There are many videos on youtube showing how focal-length changes the appearance of a subject, and there are websites with still-photo examples too. Delving into the mathematics of how apparent distance between facial features changes is way beyond my arithmetic skills!

Or you could get a camera and a few lenses and experiment...

Note: Are you asking as a "journalist" or a graphic designer?

Czardoom · Dec 9, 2021

Here's the guidance. You buy a camera and a zoom lens (or more) that covers a wide focal range. You go out and methodically take photos at various distances, with a variety of focal lengths. You do this very systematically. I think most interchangeable lens cameras record the focal length, so that will be available in the photos properties. Either study the photos carefully, or more likely use the computer to try and answer all the questions you posed.

In other words, if you want the answers to your questions, you have the capability to do so, and will get a far better answer than any of our guesses.

YuengLinger · Dec 9, 2021

Czardoom said:
Here's the guidance. You buy a camera and a zoom lens (or more) that covers a wide focal range. You go out and methodically take photos at various distances, with a variety of focal lengths. You do this very systematically. I think most interchangeable lens cameras record the focal length, so that will be available in the photos properties. Either study the photos carefully, or more likely use the computer to try and answer all the questions you posed.

In other words, if you want the answers to your questions, you have the capability to do so, and will get a far better answer than any of our guesses.

And if it's for work, the gear is a business expense! Do I hear write off? Whooo hooo!

unfocused · Dec 9, 2021

First, focal length is irrelevant. It is the distance from camera to subject that matters. If you stand 15 feet away from a subject and use a 24mm lens you will get exactly the same proportions as standing 15 feet away with a 100 mm lens. The difference is in how much space is visible around the subject. In other words if you crop the 24mm image to the same head shot of the 100 mm image they will have the exact same look and depth of field, although the final 24mm crop will have less resolution because you are discarding most of the image.

The difference comes into play because if you shoot a head shot with the 24 mm lens and have the head fill the frame of the 24mm shot you will be standing much closer to the subject. Thus the relative distance from the tip of the nose to the ears is a lot greater so the nose appears much bigger. With a 100 mm lens the relative distance from the nose to the ears is less so the nose appears smaller by comparison.

I don’t know of any uniform standards for event shots like red carpet photos of Hollywood stars. I suspect they try to set the space up so photographers can easily get both a full body shot using a 24-70 lens as well as a 3/4 length shot.

I don’t know any way to reverse engineer the shots as there are too many variables.

neuroanatomist · Dec 9, 2021

killswitch said:
I remember seeing an image where a photographer took portrait of a woman using 10-15 different focal lengths to illustrate how much distortion and flatness we can achieve using lenses from 24mm to 200mm.

This is the one I recall seeing:

But, I agree with the above – too many variables to reverse engineer this.

The distortion is a function of perspective, which is solely determined by the distance between subject and camera.

SteveC · Dec 9, 2021

neuroanatomist said:
This is the one I recall seeing:
View attachment 201594

But, I agree with the above – too many variables to reverse engineer this.

The distortion is a function of perspective, which is solely determined by the distance between subject and camera.

Based on this an 85-100 is just above the low edge of what I'd want to use. (I mention that because I own primes in those lengths.)

I have been tempted to buy the EF 135 at times, and maybe my next attack of GAS will result in that.

Joules · Dec 10, 2021

Quick edit: Obviously the perfect answer is to just look at what camera and focal length is being reported in the EXIF data of an image. The difficulty only arises if that was stripped away!

As the others have said, if you want to get the distance just from the face that's a pretty hard task to solve 'manually'. If you can also use other knowledge about the scene, like the height and distances of objects in the background, you might work it out though.

For images alone, the answer was already mentioned as well: Computers are the way to go. Here's one solution that appears to be both competent and reasonable to set up:

New version: https://github.com/sicxu/Deep3DFaceRecon_pytorch

GitHub - microsoft/Deep3DFaceReconstruction: Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set (CVPRW 2019)

Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set (CVPRW 2019) - microsoft/Deep3DFaceReconstruction

github.com

This doesn't really give you the subject distance rather than do the first step of your job for you. But given the 3D model of the face, you can of course use an image and a 3D viewer and move its camera around until the image from the 3d viewer matches the real one. Giving you the subject distance and also the focal length.

Sporgon · Dec 10, 2021

unfocused said:
First, focal length is irrelevant.

To be honest I think this is misleading when discussing perspective in photography. When taking a tight portrait we generally want to produce a flattering image of the person and this means moving away from the subject to give pleasing perspective. But then to frame that image as we want without unnecessary cropping means we have to choose a particular focal length that is long enough to frame as we want from the distance we want. So, if I want both a tight head portrait and to keep about six feet away from the subject a 200mm lens allows me to get the perspective I want. To refuse to acknowledge focal length when discussing perspective in photography is as misleading as the person who states "my 300mm gives lots of compression".

jd7 · Dec 10, 2021

Joules said:
Quick edit: Obviously the perfect answer is to just look at what camera and focal length is being reported in the EXIF data of an image. The difficulty only arises if that was stripped away!

Even if the EXIF data is there, there is a difficulty if the image has been cropped (unless you can possibly work out the crop, eg from the resolution of the image, but i don't think you could confidently do that without knowing how the image was treated in post).

privatebydesign · Dec 10, 2021

i didnt notice anybody mention sensor size in the discussion so far, which is effectively the same as cropping. If you frame the subject in the shot the same size with the same lens with a fF and crop camera the perspective changes so the subject will look different. I know this has basically been said, focal length isn’t the deciding influence. perspective (distance from the subject) is. But it’s worth bearing in mind an 85 mm lens used for headshots on a ff camera will look different, certainly to a computer assimilating many similarly framed images, to the same lens on a crop camera.

As for the lens creating the perspective, I think we put that one to rest many years ago, at least back in 2013...

No 7D Mark II in 2013? [CR2]

Let's get back to the fundamental principle here. A great shot from a A1400 from a never attempted perspective very close to an animal very difficult to do so, would destroy anything ever done by any super-tele + $$$$$ 1D combo. That's the principle. It's Irrelevant how its done, but that's...

www.canonrumors.com

jd7 · Dec 10, 2021

Sporgon said:
To be honest I think this is misleading when discussing perspective in photography. When taking a tight portrait we generally want to produce a flattering image of the person and this means moving away from the subject to give pleasing perspective. But then to frame that image as we want without unnecessary cropping means we have to choose a particular focal length that is long enough to frame as we want from the distance we want. So, if I want both a tight head portrait and to keep about six feet away from the subject a 200mm lens allows me to get the perspective I want. To refuse to acknowledge focal length when discussing perspective in photography is as misleading as the person who states "my 300mm gives lots of compression".

I don't think Unfocused was saying that focal length has no practical relevance to portrait photography, just that focal length is irrelevant to the issue the OP has asked about. Also, you say that using a 200mm lens allows you to get the perspective you want. In fact though, using a 200mm lens allows you to get the magnification and framing you want given the perspective you have separately chosen (ie about six feet). Focal length no doubt has a place in a discussion about perspective, but it should be a secondary factor in that discussion.

Kit. · Dec 10, 2021

Sporgon said:
To be honest I think this is misleading when discussing perspective in photography.

We are discussing 3D modeling here, not photography.

killswitch · Dec 13, 2021

neuroanatomist said:
This is the one I recall seeing:
View attachment 201594

But, I agree with the above – too many variables to reverse engineer this.

The distortion is a function of perspective, which is solely determined by the distance between subject and camera.

Yes, this is exactly what I remember seeing long time ago here in the forums. Thanks @neuroanatomist, this is going to my library. As someone mentioned above I can replicate this test to get a better understanding of how facial structure deforms based on the age of the subject + the focal length + distance. Distance being the tricky one. But this helps.

killswitch · Dec 13, 2021

jd7 said:
Even if the EXIF data is there, there is a difficulty if the image has been cropped (unless you can possibly work out the crop, eg from the resolution of the image, but i don't think you could confidently do that without knowing how the image was treated in post).

Yeah, typically in studios if we are provided actual copies of the photos with EXIF intact it helps. Most cases we do not have that info, so we try out best to guestimate.

killswitch · Dec 13, 2021

YuengLinger said:
For the level of precision and automation used to create believable, higher-end video games, I would guess there are programs that can analyze a series of images taken from wide-angle to super-telephoto.

There are many videos on youtube showing how focal-length changes the appearance of a subject, and there are websites with still-photo examples too. Delving into the mathematics of how apparent distance between facial features changes is way beyond my arithmetic skills!

Or you could get a camera and a few lenses and experiment...

Note: Are you asking as a "journalist" or a graphic designer?

neuroanatomist here attached the image I was referring to in my original post. I think the best way to learn and get a better understanding is replicate that test with various subjects of different age, and facial structure. Also, I am asking from a journalist point of view what is the typical lens of choice for situation mentioned in the original post. But yes, I think it will help me as an artist to learn/study this aspect by replicating the test using various focal length and adjusted distance to see how much of the jawline flattens out and or nose distorts, etc. Thanks for the post.

Kit. · Dec 13, 2021

killswitch said:
Yes, this is exactly what I remember seeing long time ago here in the forums. Thanks @neuroanatomist, this is going to my library. As someone mentioned above I can replicate this test to get a better understanding of how facial structure deforms based on the age of the subject + the focal length + distance. Distance being the tricky one. But this helps.

Let me put it this way: a lens can only bend light inside of itself. The perspective of the image is fully determined by how the light from the scene crosses the entrance pupil of the lens, so, apart from the size, shape, and position of the entrance pupil, the lens has no control over the perspective. The focal length of the lens plays no role if the distance to the subject is already known, but can in some cases be useful to calculate this distance.

Your model is still missing one or two things:
1. The mutual orientations (Euler angles or equivalent) of the head and of the lens (in addition to the distance to the lens) will determine how the 2D projection of the 3D head onto a flat sensor will look like.
2. The lens entrance pupil size (in addition to the distance to the lens) will determine how blurry the areas that are not in focus will be drawn.

stevelee · Dec 13, 2021

privatebydesign said:
i didnt notice anybody mention sensor size in the discussion so far, which is effectively the same as cropping. If you frame the subject in the shot the same size with the same lens with a fF and crop camera the perspective changes so the subject will look different. I know this has basically been said, focal length isn’t the deciding influence. perspective (distance from the subject) is. But it’s worth bearing in mind an 85 mm lens used for headshots on a ff camera will look different, certainly to a computer assimilating many similarly framed images, to the same lens on a crop camera.

Back when I was shooting a Rebel (oops, that sounds too Yankee for me), I wanted to get a lens for shooting portraits. That was before I came to this site. My first task was to decide what it was about an 85mm or 100mm lens that to me made it right for portraits. Poking around the interwebs, of course I found there was not something magical in the optics per se, but rather that framing from an optimum distance was the main thing. So I bought a 50mm f/1.4 lens, and it was exactly what I wanted.

SteveC · Dec 13, 2021

stevelee said:
Back when I was shooting a Rebel (oops, that sounds too Yankee for me),

"Shooting with a Rebel" should correct the problem and put you on the non-Yankee side.

Search

Search

Understanding Focal Length of a Portrait

killswitch

Sporgon

5% of gear used 95% of the time

YuengLinger

Print the ones you love.

Czardoom

YuengLinger

Print the ones you love.

unfocused

Photos/Photo Book Reviews: www.thecuriouseye.com

neuroanatomist

SteveC

R5

Joules

doom

GitHub - microsoft/Deep3DFaceReconstruction: Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set (CVPRW 2019)

Sporgon

5% of gear used 95% of the time

jd7

privatebydesign

No 7D Mark II in 2013? [CR2]

jd7

Kit.

killswitch

killswitch

killswitch

Kit.

stevelee

FT-QL

SteveC

R5

Similar threads