How to have your last video reviewed and then fixed

Last updated 2026-08-30 · 7 min read

What is the short version?

Asking Vizard Agent for a critique is a genuinely different request from asking for an edit. You get back a list of problems you can decide about one by one, rather than a new version built on assumptions somebody made about what you disliked.

1. Go to Vizard Agent and upload the finished video.
2. Ask for a thorough analysis of what is wrong with it.
3. Say where it is published, since the standard differs by platform.

What do you need before you start?

The published video and an open question. Vizard Agent inspects the file rather than a description of it, so you do not need to know what is wrong — that is what you are asking — but saying where it runs tells it what to measure against.

• The video. The version people actually saw.
• The platform. Phone, desktop, or both.
• Any suspicions. What felt off to you.
• Earlier videos. So the house style can be compared.
• The next step. Critique only, or critique then fix.

What do you type into Vizard Agent?

Ask for the analysis before the changes. Vizard Agent will do both, but requesting the critique first means you see the list and choose from it, instead of receiving a re-edit built around one problem it decided to prioritise.

Prompt

Variants worth knowing:

• Compared to your earlier work. Where the style drifted.
• Mobile-specific. Whether graphics survive a phone screen.
• Then the fix. The changes made after you have chosen.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real review of a published eight-minute video. The habit worth copying is in steps ten to nineteen: it sampled the earlier videos at eight separate timestamps rather than judging the style from the opening.

1. Loads the ledger, the playbook, the effects and the assets.
2. Probes the source video and checks the available tools and typefaces.
3. Transcribes the footage word for word.
4. Extracts survey and opening frames from the source.
5. Makes a working proxy so the analysis does not fight a large file.
6. Reviews the survey frames and the opening at several seconds in.
7. Samples the previous videos' style and measures the audio.
8. Reviews the earlier work at the opening, at eight seconds, at twenty and at forty-five.
9. Analyses the speech gaps, reviews the earlier style at the mid-point, at 150 seconds, late in the video and at the ending, then extracts the key word timings the graphics would need to sync to.

Step 9's spread is the point. A style that works for thirty seconds and falls apart at three minutes is invisible if you only look at the beginning, and most of the defects worth finding live in the middle where nobody checks.

What does the result look like?

The video under review measures 3840x2160, HEVC, 30fps, 483.11 seconds — eight minutes at ultra-high definition. What comes back first is a list: where the graphics fail on a phone, where the pacing sags, where the audio drops, where the style stops matching the earlier videos.

The fix follows the list rather than preceding it. That order is the whole value: you get to disagree with an item before anyone spends time acting on it.

When does this not work well?

A critique is a judgement, and a good deal of what makes a video work or fail is not measurable from the file at all. These are the limits on what Vizard Agent can tell you, and they are worth knowing before you rely on the list.

• Taste is not a defect. Vizard Agent reports problems, not preferences.
• Performance data is not in the file. Retention curves come from your analytics.
• Some faults are structural. "The premise is weak" cannot be fixed in an edit.
• Comparison needs something to compare. Without earlier videos, there is no drift to find.
• A long list is not a plan. You still choose what matters.

How do you fix a result that came back wrong?

Say which findings you accept and which you reject. Vizard Agent keeps the transcript, the survey frames, all the style samples, the audio measurements and the word timings, so the agreed fixes can be made without re-analysing the video.

• "Only fix the graphics." The other findings are left alone.
• "That is deliberate." Removed from the list and not changed.
• "You missed the audio at four minutes." Re-measured at that point.

How does Vizard Agent compare to doing it yourself?

By hand this means watching your own video again, which is the least reliable review there is — you know what you meant, so you see what you intended rather than what is on screen. Comparing it against six earlier videos is the part nobody does.

Common questions

Can Vizard Agent critique without changing anything? Yes, and it is worth doing that first. Ask for the analysis and stop there.

What kind of defects does it find? Vizard Agent reports legibility, pacing, audio levels, style drift and graphics that fail at phone size.

Does it need my earlier videos? Only for comparison. Vizard Agent finds problems in a single video perfectly well without them.

Will it just tell me it is good? No. Vizard Agent reports what it measured, including when that is uncomfortable.

Can it then fix what it found? Yes. Vizard Agent applies the fixes once you have chosen which findings to act on.

Why sample the middle rather than the opening? Because openings get the most attention when a video is made, so that is where the fewest problems survive. The faults that reach publication almost always live after the first minute.

Does it use my analytics? No. Vizard Agent reads the file. Retention data is a separate signal you hold.

Can it review somebody else's video? Yes, as a study. That is a common way to work out why a competitor's video holds up.

How long a video can it review? Any length. Longer videos simply mean Vizard Agent takes more sample points.

Should I do this before every video or after? After, and periodically rather than always. Vizard Agent finds drift by comparing across your output, so a review every handful of videos catches more than a review of each one in isolation.

Is it worth reviewing a video that performed well? Often more than one that failed. Vizard Agent measures what is actually on screen, and a video can succeed on its subject while carrying faults that quietly cost it the viewers who left early.

Does Vizard Agent check its own fixes? Yes. It reviews the corrected version against the findings before delivery.