Happy Halloween Blenders!
I am not sure where to put this question, redirect it if I chose an incorrect category.
I am trying my hand at likeness and I’m stuck.
I started gathering reference pics from the internet and made a few hundred screenshots from movies with an object of work. After I sculpted a general actor-like face I started to use cameras with projection images from movies to make the model more alike with an actor. The issue is the model looks drastically different with different Focal Length values and I don’t know the original Focal Lenght values. I sculpted the head using one sequence of images, I pick the next sequence of images and the model doesn’t fit, ok I fix the model to fit the new pics and it’s not working for the third sequence of images etc.
I read here: https://en.wikipedia.org/wiki/Normal_lens that most of the cameras in moviemaking use 35mm sensor + 35 or 50 mm lens. Another question, how do I calculate FOV of a cropped image? For example if we take the full pic as 35/35mm how do I calcualte FOV of this red square?
There will always be “distortions” in a 2D representation of a 3D object (a person), that’s an inherent sacrifice of removing a dimension. The best you can do is work with those distortions and understand how they would relate to the 3D object (the person)
Also bare in mind that often lenses are much farther away from the actor than you think, and often zoom in for a closer shot. This is why red carpet images from Getty Images and the like are often excellent reference material as they use super long lenses quite far away, which can give an almost orthographic look.
There’s no real ‘good’ answer here, other than the age old ‘experience’. But really, it’s just about getting the feel/eye for it. For something like above, try something higher, like 120-200 range.
Also, use your 3D cursor as the pivot point, and set it on your model when aligning the camera.
Actions such as G + Z+Z (with camera selected) allows you to essentially zoom in and out without changing the overall position of the camera, and is very useful for changes to FOV.
Also, use the eyes and the ears when aligning. Try a value (and it’ll always be higher than you think) and then align it with the eyes, and the ears if possible, which will give you the right angle. Then tweak fov and adjust camera position until its closer. Don’t worry if something doesn’t quite align properly. It’s almost impossible to manually get multiples different images to align perfectly like this. Just look for the major landmarks and forms. (Eyes, nose, mouth, and oveall facial shape/structure.) There is always going to be some artistic lisence, and you shouldn’t rely in this method alone, and should do some sculpting from observation aswell.
You’re already off on a path that will lead confusion. Not your fault. That sentence over-generalizes quite a bit. It works well for a layman’s explanation (i guess), but not so well for a technical audience.
Firstly: FOV is simply an angle, nothing else. It is the angle of view that was visible to the camera (in digital, visible to the camera sensor).
Now, the combination of the sensor size and the focal length of the lens is what gives you the FOV result.
As to the sizes - "most of the cameras in moviemaking use 35mm sensor… S35 is a highly used sensor standard, but there’s also LF and FF. For Game of Thrones, S35 is probably your choice based on the cameras they usually used. I think Alexa may have an actual sensor size of something like 34.5(?) (or .9? can’t remember), but it still counts.
“+ 35 or 50 mm lens”
Those sizes ARE used, but other sizes are used at well. Often a prime lens, but sometimes a zoom might be called for. 35 & 50 are good numbers to try and get into the basic range, though. For some shots you can generally sort of guess a basic focal length, based on how flat the image looks and how blurry the DOF.
Now to your basic question: how to you determine the actual FOV, if you were to be given a photo of Daenerys cropped to that red square?
Assuming the FOV is not in the metadata of the photo file, I’m not sure that you could. If I had to, I’d be setting up a camera in real life with some lenses, making some educated guesses (as listed above), and see how close I could get.
In Some movies the cinematographer choose to only use one lens (and most of the time it’s a 35 on a 50) but in most of movies, there is many different lens used.
Sometime they use very wide angle (28 or even 20) for different reasons (wide view of a landscape, exaggeration of perspective, different feeling of camera movement…) or long focal as 85 or 135 (change perception of distance, less optical distortion to have more beautiful face…)
Find exact focal used in a movie is really hard, but with a bit of experience, you can easily know if it’s wide or long focal.
