Matching colour across cameras

Two cameras, same table, same lamp, same afternoon, and one of them thinks the room is warm while the other thinks it is grey. Cut between them and your video appears to move house every few seconds.

The instinct is to grade each camera until it looks nice. That is the wrong shape of job, and it gets slower with every angle you add, because you are chasing a look you cannot hold in your head while you switch between three timelines.

Do it the other way round. Pick the one angle you trust, make that one correct, and then match every other camera to it rather than to your idea of what good looks like. One reference, one decision, and the rest is comparison. That is the whole method and everything below is how to make it work on your own footage.

Why do two cameras in the same room look different?

Three reasons, and only one of them is the camera's fault.

Every manufacturer makes its own decisions about how to render colour, so two bodies see the same wall and disagree politely. Cameras also sit at different angles to your light, so one is getting the lamp full on and the other is getting it across the face, which changes both brightness and colour before anything is recorded. Most importantly, automatic white balance is doing its own thing in each camera independently, and it does not stop doing it when you press record.

That last one is the reason the problem often gets worse mid take rather than staying still.

Should you white balance manually?

Yes, and this is the single most useful thing in this post.

Automatic white balance can change during a take as the scene changes. That drift gives the grade more to chase than one stable offset. Where the setup allows it, set and check white balance before recording.

Set white balance manually on every camera before you start and leave it there. Most cameras let you pick a preset or dial in a value, and either is fine. What matters is that it stops moving. If you can, set the same value on every body, even if that value is not perfect, because one matched error is far easier to correct later than two unmatched approximations. That settles the warm to cool axis, not the green to magenta one: two bodies on the same value can still disagree there, because manufacturers render that part differently, which is why the reference clip below matters more than the number on the dial.

The reference clip

A few seconds of the same subject on every camera, under the lighting you are about to use.

Sit where you will sit. Roll all cameras. Hold something with a known colour in front of you, a grey card if you own one and a sheet of white paper if you do not, and let each camera see it for a moment. The card is the better reference: the brighteners in most white paper fluoresce faintly blue, so it reads cooler than neutral, and by a different amount under each light. Then say what you are doing out loud so you can find it later.

That handful of seconds is your reference. When the footage is on the timeline you have every camera looking at the same object under the same light, which turns matching from a taste exercise into a comparison. Without it you are matching a shot of your face to a shot of your hands and guessing at the difference between the two.

What does matching to a reference mean?

Pick a useful reference angle and get its exposure, neutral tones and skin where they need to be. That gives the other shots something concrete to match. The rest still needs judgement; it no longer starts from a blank screen.

Then bring each other camera up beside it and adjust that camera until its reference clip matches the reference clip of the trusted angle. Not until it looks nice on its own. Until it matches. Your editing software will have a way of showing two frames side by side, and your eyes are considerably better at spotting a difference between two images than at judging one image alone.

Use the reference correction as a starting point for that camera, then watch the actual cuts. A light change, exposure shift or different part of the scene may need a local adjustment. Keep the room consistent; do not mistake identical settings for matching pictures.

This is how this studio works on multi-camera pieces: one reference angle holds the truth, and everything is matched to it. A three camera lesson has to feel like one room, and it only does if one of the three has been appointed the room.

What if the cameras are lit differently?

Then fix the light, not the file.

What matters is how the subject looks from each angle. A face and an overhead worktop may receive very different light. Check both before recording; sometimes moving or softening a lamp is simpler than forcing a match afterwards.

If the lighting cannot be changed, ask what the difficult angle shows. A slightly poorer colour match may be worth keeping for a clear view of the action. If another angle shows the same information just as well, use that instead.

Where this stops being a colour problem

Manual settings, a shared reference and considered lighting give the grade a better start. The final check is still the cut: do these pictures belong in the same room?

Spend the first moments of the next shoot giving the cameras something to agree on: the same reference, in the light you will use. Later, you can put those frames side by side instead of asking your memory what the room looked like.

Sources and evidence

What here is established colour science and what is our own method.

Colour context. Camera-specific transforms account for differences in recorded colour data. Daylight itself also varies; a daylight locus describes a model of its chromaticities rather than a fixed path through every day. A shared reference helps compare cameras, while the final match remains a production judgement.

Our method. Filming a shared reference clip before anything else, and matching every other camera to the one trusted angle rather than to its own idea of good, is the practice here.

References.

• Academy / ACES. Design, Integration and Use of ACES Look Transforms. docs.acescentral.com. A Look Transform is a defined, portable transform on scene-referred ACES values, separate from creative grading. Grading itself is outside ACES scope; does not define any particular creative look.
• ARRI. Log C. arri.com. Log C is a scene-referred log encoding preserving captured dynamic range for grading; not display-ready without a transform. Log does not add dynamic range or look correct without a transform.