Vizard Agent

How to make the background colour and lighting consistent in a video

Last updated 2026-08-16 · 6 min read

Upload the clip to Vizard Agent, say which part of the picture looks uneven, and say what must be left alone. Vizard Agent measures which pixels move across the whole clip before touching anything, isolates the subject from the background, evens out the background's colour and lighting, and then checks that the subject's pixels are unchanged.

What is the short version?

Patchy ground, a wall lit unevenly, a sky that shifts halfway through — these read as amateurish even to people who cannot say why. The fix is easy to describe and hard to do by hand, because any adjustment strong enough to even out the background also lands on the thing in front of it.

  1. Go to Vizard Agent and upload the clip.
  2. Say which area looks uneven: the ground, the wall, the sky, the lighting.
  3. Say plainly what must stay exactly as it is.

What do you need before you start?

The clip and a clear statement of what is protected. Vizard Agent works out the rest by measurement, and the one thing it cannot infer is which part of the picture you consider untouchable, which is exactly the part a colour adjustment is most likely to damage.

What do you type into Vizard Agent?

Say the area and the exception in the same sentence. Vizard Agent treats the protected thing as a hard constraint rather than a preference, and naming it explicitly is what turns a general colour pass into a targeted one.

Prompt

Even out the colour, brightness and lighting of [the ground / the background] so it looks consistent across the clip. Keep [the subject] exactly as it is now.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real background-matching job. Almost half the run is measurement before a single pixel changes, because the question "what is safe to adjust" has an answer that can be computed rather than guessed.

  1. Probes the clip and extracts a grid of frames to see the problem.
  2. Pulls key frames at full resolution rather than judging from thumbnails.
  3. Analyses the camera motion and stability across the clip.
  4. Differences the frames against each other to find what changes.
  5. Builds a motion-variance map showing which pixels move and which are static, then looks at it.
  6. Crops the subject at several points in time and tiles them into one comparison grid.
  7. Measures colour, brightness and contrast statistics for each region of the background.
  8. Harmonises the background with an image edit, then resizes the result back to the clip's exact dimensions.
  9. Compares the original and the harmonised frame side by side, and checks the subject's pixels for any difference.

Step 5 is the one worth knowing about. Vizard Agent found the boundary between subject and background by measuring movement over time rather than by drawing a mask by eye, and step 9 is it proving to itself that the protected area survived.

What does the result look like?

From the run this page is written from, probed on the delivered file: 540x960, H.264, 30fps, 5.3 seconds, AAC audio. A short vertical clip delivered at the source's own dimensions, because a colour job should not quietly change the resolution of what you sent.

Ask for a larger output and Vizard Agent will scale, though matching the source is the sensible default when the point of the job is that nothing else changed.

This category was not separately measured, so treat the timing as a range rather than a promise: comparable work runs a median of 28 to 38 minutes end to end. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

Evening out a background means inventing a version of it that was never filmed, and the more of the frame that has to be invented the more likely it is to show. Vizard Agent measures carefully and works within what the footage supports, and no measurement makes a badly lit shot into a well lit one.

How do you fix a result that came back wrong?

Say which way it went wrong. Vizard Agent keeps the motion map, the measurements and the isolated subject, so backing off the strength or protecting a further area works from the analysis it already built rather than starting the measuring over.

How does Vizard Agent compare to doing it yourself?

By hand this is a mask, a tracker, a colour node, and an afternoon of watching the edge crawl. The tools are excellent and the work is fiddly, which is why most people either skip the fix or apply a grade to the whole frame and accept the collateral damage.

By hand Vizard Agent
Finding the subject's edge Rotoscope or key it Measured from motion across time
Deciding the correction Eyeball the histogram Statistics per region
Protecting the subject Mask and hope Verified pixel-for-pixel
Adjusting later Re-track, re-render Say which way

Common questions

Will it change my product or subject? No, when you say not to. Vizard Agent isolates it and checks afterwards that those pixels are unchanged.

Can it match two different clips to each other? Yes. Say which one is the reference and Vizard Agent measures both.

Does it work on a moving camera? It analyses camera motion first. A static or gently moving shot is where this works best.

Will the video stay the same size? Yes. Vizard Agent delivers at the source dimensions unless you ask for something else.

Can it even out the sky as well as the ground? Yes. Name the region and Vizard Agent measures that area specifically.

Is this the same as colour grading? No. Grading sets a look for the whole picture; this evens out one region while leaving everything else alone.

How long does a clip like this take? Most of the time goes on measurement rather than rendering, and Vizard Agent front-loads that work so later adjustments come back quickly.