Why You Can’t Re-light Footage in Post

The concept of “re-lighting in post” is one that has enjoyed a popularity amongst some no-budget filmmakers, and which sometimes gets bandied around on much bigger sets as well. If there isn’t the time, the money or perhaps simply the will to light a scene well on the day, the flexibility of RAW recording and the power of modern grading software mean that the lighting can be completely changed in postproduction, so the idea goes.

I can understand why it’s attractive. Lighting equipment can be expensive, and setting it up and finessing it is one of the biggest consumers of time on any set. The time of a single wizard colourist can seem appealingly cost-effective – especially on an unpaid, no-budget production! – compared with the money pit that is a crew, cast, location, catering, etc, etc. Delaying the pain until a little further down the line can seem like a no-brainer.

There’s just one problem: re-lighting footage is fundamentally impossible. To even talk about “re-lighting” footage demonstrates a complete misunderstanding of what photographing a film actually is.

This video, captured at a trillion frames per second, shows the tranmission and reflection of light.

The word “photography” comes from Greek, meaning “drawing with light”. This is not just an excuse for pompous DPs to compare themselves with the great artists of the past as they “paint with light”; it is a concise explanation of what a camera does.

A camera can’t record a face. It can’t record a room, or a landscape, or an animal, or objects of any kind. The only thing a camera can record is light. All photographs and videos are patterns of light which the viewer’s brain reverse-engineers into a three-dimensional scene, just as our brains reverse-engineer the patterns of light on the retinae every moment of every day, to make sense of our surroundings.

The light from this object gets gradually brighter then gradually darker again – therefore it is a curved surface. There is light on the top of that nose but not on the underneath, so it must be sticking out. These oval surfaces are absorbing all the red and blue light and reflecting only green, so it must be plant life. Such are the deductions made continuously by the brain’s visual centre.

A compound lens for a prototype light-field camera by Adobe

To suggest that footage can be re-lit is to suggest that recorded light can somehow be separated from the underlying physical objects off which that light reflected. Now of course that is within the realms of today’s technology; you could analyse a filmed scene and build a virtual 3D model of it to match the footage. Then you could “re-light” this recreated scene, but it would be a hell of a lot of work and would, at best, occupy the Uncanny Valley.

Some day, perhaps some day quite soon, artificial intelligence will be clever enough to do this for us. Feed in a 2D video and the computer will analyse the parallax and light shading to build a moving 3D model to match it, allowing a complete change of lighting and indeed composition.

Volumetric capture is already a functioning technology, currently using a mix of infrared and visible-light cameras in an environment lit as flatly as possible for maximum information – like log footage pushed to its inevitable conclusion. By surrounding the subject with cameras, a moving 3D image results.

Sir David Attenborough getting his volume captured by Microsoft

Such rigs are a type of light-field imaging, a technology that reared its head a few years ago in the form of Lytro, with viral videos showing how depth of field and even camera angle (to a limited extent) could be altered with this seemingly magical system. But even Lytro was capturing light, albeit it in a way that allowed for much more digital manipulation.

Perhaps movies will eventually be captured with some kind of Radar-type technology, bouncing electromagnetic waves outside the visible spectrum off the sets and actors to build a moving 3D model. At that point the need for light will have been completely eliminated from the production process, and the job of the director of photography will be purely a postproduction one.

While I suspect most DPs would prefer to be on a physical set than hunched over a computer, we would certainly make the transition if that was the only way to retain meaningful authorship of the image. After all, most of us are already keen to attend grading sessions to ensure our vision survives postproduction.

The Lytro Illum 2015 CP+ by Morio – own work, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=38422894

But for the moment at least, lighting must be done on set; re-lighting after the fact is just not possible in any practical way. This is not to take away from the amazing things that a skilled colourist can do, but the vignettes, the split-toning, the power windows, the masking and the tracking – these are adjustments of emphasis.

A soft shadow can be added, but without 3D modelling it can never fall and move as a real shadow would. A face can be brightened, but the quality of light falling on it can’t be changed from soft to hard. The angle of that light can’t be altered. Cinematographers refer to a key-light as the “modelling” light for a reason: because it defines the 3D model which your brain reverse-engineers when it sees the image.

So if you’re ever tempted to leave the job of lighting to postproduction, remember that your footage is literally made of light. If you don’t take the time to get your lighting right, you might as well not have any footage at all.

Why You Can’t Re-light Footage in Post