Cinematics Sans Cutscenes

Close-Ups

Ever since first witnessing the sniper-esque zoom in Quake, I’d wondered why enemies viewed that way looked much more realistic and solid than in the regular view. Later as a game developer I learned it’s because while first-person camera field-of-views tend to require a wide angle (around 70-90 degrees) to emulate the sense of environmental awareness afforded by a human eye on a computer or television screen, the eye in fact only focuses on an angle of 30 degrees directly in front of them, making the world we see appear the same as if viewed by a videogame sniper scope, with the remaining visible 150 degrees offering  merely peripheral vision.

While film cameras can range anywhere between 180 (fisheye) and <1 degrees (super telephoto), they tend to stay around 40 degrees (36mm), better emulating the human eye and causing any use of similar cameras in our games to be described as “cinematic”. Below is a comparision of a character viewed with both wide and narrow camera fields of view respectively, with the former looking familiar to anyone who has ever gotten close enough to an NPC in a 1st-person game, and the latter similar to a typical screenshot from a conversation in Mass Effect.

Employing field-of-views similar to the second example goes some way to making the character appear more solid and life-like due to it’s close proximity to how we view others in the real world with our own eyes as opposed to the warped “game-like” character in the first image.

Below is another example from God of War, this time showing how affecting the field-of-view in gameplay is not distracting and can be used to punctuate events without disorientating the player or giving the same feeling of loss of control as when affecting rotation.

Beyond better framing actions, affecting the field-of-view in this manner allows us to really get into new territory previously unexplored by videogames until the recent advances in character fidelity. Going close on a character would previously have only served to highlight the low polygon counts, lack of movement and variety in facial features and dull, lifeless eyes, but now zooming in allows us to get into a deeper world of subtlety in body language and character acting.

Below is an example from the original Mass Effect 1 – one of hundreds of procedurally-generated lines of conversation using no animator time beyond an initial setup of expressions, gestures and cameras and the appropriate rules and algorithms to decide when to use them. Notice that beyond getting in close enough to see minute facial movements, we are framing the characters in a manner akin to cinema, using long lenses (narrow field of view) to avoid the aforementioned game-like aesthetics provided by the closer, wide-angle cameras used for gameplay.

Impotantly, the ever-present videogame problem of cameras embedding themselves in walls is overcome here by, rather than ensuring the cameras themselves are safe from obstruction, instead ensuring only the near-clip planes are of concern and must be placed correctly (note that the alien Wrex is standing with his back to a crate, but we can still have a shot over his shoulder despite the camera being placed well behind the crate and inside a wall – a technique akin to moving the set around on a film shoot to enable the desired shots).

While the above conversation is set up with pre-defined positioning of characters (controlled subjects), affecting the field-of-view dynamically based on the distance between protagonists would go some way to filling the frame despite our hypothetical uncontrolled cameraman, who can move or be placed randomly when initiating any cinematic sequence.

All this dynamic framing of subjects, however, means nothing unless we can also dynamically compose our shots too, but this issue has been dealt with a long time ago in our medium, with refinement occurring in the preceeding years, which leads us to our final cinematic tenet that we must emulate.

But first, to summarise Close-Ups:

  • Close-ups are essential for character performances.
  • A dynamic field-of-view enables the filling of the frame.