How to cut a trailer for an online course from a script

Last updated 2026-08-23 · 6 min read

What is the short version?

A course trailer needs the presenter saying the sales script, and the presenter recorded something else six months ago. Rather than reshooting, the footage can be made to deliver the new words — which only works if the timing and the lip movement are handled properly.

1. Go to Vizard Agent and give it the script.
2. Upload the footage of the presenter you already have.
3. Say the length, the shape and the tone.

What do you need before you start?

The script, the footage and permission. Vizard Agent handles the voice, the timing and the picture, and the one thing that must be settled before you start is that the person on screen has agreed to their voice and likeness being used this way.

• The script. The exact words the trailer should say.
• The footage. Any usable take of the presenter.
• Permission. Explicit, from the person being cloned.
• The length. Thirty seconds for a trailer.
• The end card. A URL, an offer, a call to action.

What do you type into Vizard Agent?

Paste the script and say what it is for. Vizard Agent works out the pacing, the atmosphere shots and the grade from the words themselves, so the script is doing most of the briefing and you rarely need to describe the look separately.

Prompt

Variants worth knowing:

• Generated atmosphere shots. For the beats where the presenter is not on screen.
• A dark grade. It suits this category and Vizard Agent will check it on frames.
• An upscale. If the source footage is soft, say so.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real course trailer. The chain in the middle is the interesting one: clone the voice, re-time it to real pauses, then make the existing footage match the new read.

1. Inspects the footage and measures the content area within the frame.
2. Transcribes the existing footage and checks the available tools.
3. Generates the voice clone, the music and the atmosphere shots in one pass.
4. Measures the voiceover against the clips and reviews the generated b-roll.
5. Reads the word timings and the music beats, then finds the real pauses in the narration.
6. Re-times the narration to those pauses and builds the title graphics.
7. Lip-syncs the presenter's footage to the new narration and checks the result on frame.
8. Locates and verifies the face position, and runs an upscale on the plate in the background.
9. Rebuilds the cut on the upscaled plate, locks the segment lengths to the timeline, adjusts the dark grade after reviewing frames, builds a music duck envelope and re-renders.

Step 5 is the small decision that makes the read sound human. A generated narration has even spacing; finding where the real pauses fall and re-timing to them is what stops it sounding like a machine reading a list.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1920x1080, H.264, 25fps, 30.0 seconds, AAC audio. Widescreen, exactly thirty seconds, the presenter delivering the new script with matched lip movement, generated atmosphere shots between beats, animated titles and an end card.

Exactly thirty seconds happens because the segment lengths were locked to the timeline rather than trimmed at the end. A trailer that runs to a slot length has to be built to it, and the narration timing drives everything upstream of that.

When does this not work well?

Cloning a voice and changing what someone appears to say on camera is a powerful thing to do and it carries real obligations. Vizard Agent will do the technical work; whether it should be done at all is a question settled before the job starts.

• Permission is not optional. Never clone a voice or lip-sync a face without explicit consent.
• Do not put words in anyone's mouth. The script should be theirs or approved by them.
• Lip-sync has limits. Extreme angles and fast head movement resist it.
• Upscaling recovers sharpness, not detail. A soft source stays a soft source underneath.
• A trailer is a promise. What it claims about the course is a marketing responsibility.

How do you fix a result that came back wrong?

Say which line or shot. Vizard Agent keeps the cloned narration, the word timings, the lip-synced plate, the generated shots and the grade tests, so a script change is a re-record and a re-sync rather than a rebuild from the footage.

• "Change that line." Re-recorded in the same voice and re-synced.
• "The grade is too dark." Adjusted and checked on frames, as this run did.
• "The end card URL is wrong." Re-rendered and repositioned.

How does Vizard Agent compare to doing it yourself?

By hand this means booking the presenter for a reshoot, waiting weeks for a diary gap, and then grading, upscaling and cutting the result — or giving up on it and putting a voiceover over stock footage instead.

Common questions

Do I need new footage of the presenter? No. That is the point — existing footage is made to deliver the new script.

Is the voice really his? It is a clone of his voice, which requires his explicit permission to make or use.

Will the lip movement match? Vizard Agent lip-syncs the plate to the new narration and checks it on frame.

Can it fix soft footage? It upscales the plate, which recovers sharpness rather than inventing detail.

Does Vizard Agent time the narration? Yes. It finds the real pauses in the read and re-times the narration to them, which is what stops a generated voice sounding evenly spaced.

Does Vizard Agent check the grade? Yes. It reviews frames from the cut and adjusted the dark grade twice on this run before rendering.

Does it add b-roll? Yes, generated atmosphere shots for the beats where the presenter is not on screen.

Does this work for other kinds of trailer? Yes. Any promotional cut where the script exists and the footage does not say it — a webinar, a book, a service — follows the same chain.

Can Vizard Agent generate the shots I do not have? Yes. Vizard Agent generated the atmosphere shots on this run for the beats where the presenter was not on camera.

Does Vizard Agent balance the mix? Yes. Vizard Agent measures the narration against the music and builds a duck envelope so the voice sits on top throughout.

Does it hit an exact length? Yes. The segment lengths are locked to the timeline rather than trimmed at the end.