Vizard Agent

How to turn raw footage into an Amazon Influencer product review

Last updated 2026-08-19 · 6 min read

Upload your takes and tell Vizard Agent to open with your introduction and brand mention. Vizard Agent transcribes everything, cuts the review from the transcript, re-transcribes the assembled cut to catch false starts it left in, and captions it with the brand name spelled correctly.

What is the short version?

A product review recorded in one long take is mostly usable and not in one piece. The edit that sells is the same person saying the same things with the restarts, the pauses and the second attempts taken out, and b-roll over the moments where hands do the talking.

  1. Go to Vizard Agent and upload every take, including the short ones.
  2. Say it should open with your introduction and brand shoutout.
  3. Say the length and that you want captions.

What do you need before you start?

Your footage and the opening. Vizard Agent finds the structure by reading the transcript, so it does not need you to log the take, and the two things worth stating are how the review should begin and how long the finished piece should run.

What do you type into Vizard Agent?

Say what the opening has to contain. Vizard Agent will otherwise choose the strongest hook it finds in the transcript, which is usually right for social and wrong for a review format where the introduction and the brand mention are the point.

Prompt

Turn my raw footage into an engaging [Amazon Influencer] product review. Start with my introduction and brand shoutout, at least [3-4] minutes, with captions.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real product review. The pattern to notice is that it does not trust its own first assembly: after cutting, it transcribes the cut again to find what slipped through.

  1. Checks the uploads and transcribes all four clips at once.
  2. Reads the full transcript, including the thirty-two-minute take.
  3. Builds contact sheets of the short clips and the long take and reviews them.
  4. Checks the frames at the candidate cut points and pulls full-size frames to judge the framing.
  5. Pulls exact word timings for the chosen sections and lays out the segment list.
  6. Tests the reframing crops, cuts the selected takes and joins them.
  7. Transcribes the assembled cut and re-cuts with the false starts removed.
  8. Picks the b-roll in-points and locates the caption cues for the cutaways.
  9. Tests caption styles at size, compares two candidates, adds a shadow for readability, fixes the brand spelling and renders the final cut.

Step 7 is the habit that separates this from a rough assembly. Reading the joined cut back as text surfaces the "sorry, let me start again" that survived the first pass, and Vizard Agent removes it before anyone hears it.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1920x1080, H.264, 29.97fps, 316.59 seconds, AAC audio. Widescreen, five minutes and sixteen seconds, cut from a thirty-two-minute take plus three short clips, captioned and with b-roll over the hands-on moments.

The delivery running past the requested four minutes is Vizard Agent choosing content over the brief. The material supported it, and a review that cuts a genuine demonstration short to hit a number is worse for the thing it exists to do.

When does this not work well?

A review is a claim about a product, and Vizard Agent is cutting your claims together rather than checking whether they are true. Everything to do with accuracy, affiliate disclosure and the platform's own rules stays on your side of the job.

How do you fix a result that came back wrong?

Quote the line. Vizard Agent keeps every transcript, the word timings, the tested caption styles and the b-roll in-points, so removing a section or swapping a cutaway is a targeted re-render rather than a fresh pass over half an hour of footage.

How does Vizard Agent compare to doing it yourself?

By hand this is scrubbing a thirty-two-minute take for the good version of each sentence, cutting them together, and then watching the whole thing back to find the three restarts you missed on the first pass. Vizard Agent reads the assembly as text instead, which is faster and catches more.

By hand Vizard Agent
Finding the good takes Scrub and remember Read from the full transcript
False starts Catch them on rewatch Cut re-transcribed and cleaned
B-roll placement Eyeball it Cued against the caption timings
Caption legibility Set it once Styles tested and compared at size

Common questions

How long should a product review be? Three to five minutes. Long enough to demonstrate, short enough to finish.

Will it keep my introduction? Yes, if you say so. Otherwise Vizard Agent leads with the strongest hook.

Does it add captions? Yes, and it tests the style at full size before committing to one.

What if the brand name is unusual? Write it out in your instruction. Vizard Agent will use your spelling in the captions.

Can it use my own b-roll? Yes. Upload the close-ups and Vizard Agent places them over the matching claims.

Can I get a vertical cut as well? Yes. Ask for both and the short vertical version comes from the same transcript.

Does it check the audio levels? Yes. Vizard Agent measures loudness at several points and again on the final render rather than trusting the first mix.

Will it keep the whole demonstration? If the material supports it. On this run Vizard Agent ran past the requested length rather than cutting a demonstration short.

How many takes can it handle? As many as you upload. Vizard Agent transcribes them all at once before deciding what to use.

Can it reframe the shots? Yes. Vizard Agent tests the candidate crops and checks the framing on full-size frames before it applies any of them to the cut.