Vizard Agent

How to fix pale, flat-looking footage of a person

Last updated 2026-08-16 · 5 min read

Upload the recording to Vizard Agent and say what looks wrong. Vizard Agent samples frames into a grid, tests several levels of colour correction on those stills, compares them side by side, and only then applies the one that works to the whole video. Sharpening is tested the same way.

What is the short version?

Most footage of people that "looks bad" is not badly shot. It is under-lit and flat, which reads as pale skin and soft detail. Both are correctable, and the correction is a narrow range: a little too far and the person turns orange, which is worse than pale.

  1. Go to Vizard Agent and upload the recording.
  2. Say what is wrong in plain words: pale, flat, soft, too dark.
  3. Say what should not change about how the person looks.

What do you need before you start?

The recording as it came off the camera, and a description of the problem rather than a prescription for the fix. Vizard Agent measures the picture and tests options, so it does not need you to know the settings — what it needs is which direction bothers you.

What do you type into Vizard Agent?

Describe the symptom rather than the remedy. Vizard Agent works the numbers out by testing them on your actual frames, so a prompt that names the problem gets further than one guessing at saturation values — and if you name a value, Vizard Agent will use it even when its own tests suggest otherwise.

Prompt

Make this video look better. Cut the parts that are not needed, and fix the colour so the person does not look pale — natural, not heavy, and sharper.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real recording that came back looking washed out. The middle of the run is the method that matters: it does not guess a correction and apply it, it makes several and compares them.

  1. Analyses the file for resolution, duration and audio format.
  2. Transcribes the speech, which is how it decides what can be cut.
  3. Extracts frames to see how the person actually looks.
  4. Builds those frames into a grid so the whole video can be judged at once.
  5. Extracts nine selected frames into a 3x3 grid and reads it.
  6. Analyses movement and activity to decide which sections to remove.
  7. Tests several levels of colour correction — saturation, contrast and brightness — on stills.
  8. Reads the test image to compare the options against each other.
  9. Tests sharpening levels the same way before applying any.

Steps 7 to 9 are the whole reason this works. Testing a correction on a still costs nothing; testing it on a video costs a render, which is why people who do this by hand tend to try one setting and live with it.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 25fps, 22.84 seconds, AAC audio. The source resolution and frame rate were kept, and the video came back shorter than it went in because unnecessary sections were trimmed as part of the same job.

Ask for correction without trimming and Vizard Agent keeps the full length. The two are separate instructions even though they arrive in one sentence.

How long it takes was not measured separately here. Comparable work runs a median of 28 to 38 minutes end to end, and the upper quartile is several times the median. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

Correction redistributes what the camera recorded. Where the light was wrong at the time, there is a limit to how much of it can be repaired afterwards, and pushing past that limit looks worse than the original problem.

How do you fix a result that came back wrong?

Say which way it went wrong, in one word. Vizard Agent keeps the test grid, the settings it chose and the transcript it made, so an adjustment moves the correction that already exists rather than starting the analysis from the beginning — which is why a second pass here is quick and cheap.

How does Vizard Agent compare to doing it yourself?

The manual version is a filter, applied at whatever strength looked right in the first ten seconds. Doing it properly means scopes, a test render, watching it on a real screen, and adjusting — which is why almost nobody does it properly on a talking-head video.

A filter Vizard Agent
Choosing the correction One preset, applied whole Several tested on stills
Judging it The first ten seconds A grid across the video
Sharpening On or off Levels tested before applying
Adjusting Change the preset, re-export Say which way

Common questions

Will it change how I look? It changes light, colour and clarity. It does not reshape features, and if you want it kept conservative, say so.

Can it fix bad lighting? Within limits. It can lift and balance what was recorded; it cannot put a light where there was none.

Will it trim my video too? Only if you ask. Say "do not cut anything" if you want correction alone.

Does it re-encode the whole file? Yes, the video is written fresh, though Vizard Agent keeps the source resolution and frame rate unless you ask for something different.

Can it match another video of mine? Yes. Upload the one that looks right and Vizard Agent matches it rather than deciding on its own.