How to make a silent demo video for a presenter to talk over live
Tell Vizard Agent that a live presenter will speak over it. That single fact changes the pacing: every hold has to last as long as the presenter's sentence about that screen, not as long as a viewer needs to read it, and the file has to be genuinely silent rather than merely without music.
What is the short version?
You are building the thing that plays on the screen behind someone at a demo or a webinar. It carries no sound of its own, it must not race ahead of what they are saying, and it still has to finish inside the slot. Vizard Agent paces it to the speaker.
- Tell Vizard Agent a presenter narrates it live.
- Give it the runtime ceiling as a hard limit.
- Check the delivered file contains no audio stream at all.
What do you need before you start?
The screen recordings and the running order. A tutorial like this is a sequence of workflows, and naming them in order — with the title card text for each — is what turns a pile of captures into a structure.
Give it the brand too, and any earlier videos in the series. In the session behind this article the site's actual typefaces and its signature orange were pulled from the website, and the chapter cards were matched to the format of three previous tutorials, so the new one sat in the set rather than beside it.
What do you type into Vizard Agent?
Say the narration is live, and say the ceiling. Those two constraints pull against each other — longer holds against a fixed runtime — and naming both lets Vizard Agent make that trade deliberately rather than discovering it at the end.
There is no sound and no music. A live presenter explains what is on screen, so pause for emphasis and give them time, but keep it under four minutes.
That is close to the real brief. The tension in it is the whole job: the first version of a cut like this is either too fast to talk over or too long for the slot.
What does Vizard Agent actually do?
It finds the moments that need a hold and takes the time back from the moments that do not. Loading screens, cursor travel and half-typed input are dead weight in a presenter-paced video; completed results are where the speaker has something to say.
In that session the shape of the work was:
- Identified each workflow in the recordings and the result screen that ends it.
- Pulled the website's real typefaces and colour, and matched the chapter-card format of the earlier videos in the series.
- Cut out the loading screens while preserving every key action.
- Added steady reading holds before each chapter advance, and longer ones on comparisons, forecasts and export confirmations.
- Kept every description outside the software interface so the product itself was never covered.
- Assembled at three minutes fifty and confirmed the file contained no audio stream.
Then it fixed the pacing problem you only see on playback: the search examples moved on while the question was still being typed. The correction held each prompt until it was fully typed and readable, without the runtime moving.
What does the result look like?
A video someone can talk over comfortably, without racing them. It lands on each finished result and waits there, the chapter cards give the presenter a beat to change topic, and nothing important in the interface is hidden behind an overlay.
It is also genuinely silent — no audio stream, not a muted one. That matters at an event, where a file with a silent track can still take over the room's audio path and interrupt a live microphone feed.
When does this not work well?
When the runtime ceiling is too tight for the content. Presenter pacing costs seconds per screen, and past a certain number of workflows the honest answer is fewer chapters rather than faster holds. Vizard Agent will say which chapters it had to rush.
It also depends on knowing the presenter's pace, which you know and it does not. If they speak slowly or take questions, tell Vizard Agent to lengthen the holds beyond the default, because the default is a reading speed.
And a recording where the interface itself is the problem — tiny type, a cluttered screen — will not be fixed by pacing. That needs a zoom, and a zoom needs room in the runtime.
How do you fix a result that came back wrong?
If it races ahead of you, name the chapter rather than asking for the whole thing to be slower. Vizard Agent extends that particular hold and takes the seconds back from a loading screen somewhere else, so the runtime ceiling survives the change.
If a screen is unreadable, say whether it is the size or the duration. Those have different fixes — a zoom or a hold — and asking for "longer" when the type is too small produces a longer unreadable shot.
If something appears mid-action, that is the typing problem. Ask Vizard Agent to hold on the completed state rather than on the moment the click lands, which is a different frame than it sounds.
How does Vizard Agent compare to doing it yourself?
Editing this by hand means guessing at holds, and the guess is always calibrated to yourself watching it alone, which is much faster than a presenter talking. Most of these videos get re-cut after the first rehearsal for exactly that reason.
The difference is treating the finished results as the anchors and the loading screens as the budget. Working out where every workflow actually completes, holding there, and taking the time back from the dead passages is what lets a presenter-paced cut still fit the slot — and it is arithmetic rather than instinct.
Common questions
Why must the file have no audio at all? A silent track can still claim the room's audio path. Vizard Agent verifies there is no stream.
How long should a hold be? Long enough for one or two sentences. Tell Vizard Agent if your presenter is slower.
Can it add captions instead? It can, but not for this. On-screen text competes with a live speaker.
What about the chapter cards? They give the presenter a beat to change topic. Vizard Agent times them for that.
Will it match our previous videos? Yes. Send them and Vizard Agent matches the card format and typography.
Can it pull our brand colours from the site? Yes. Vizard Agent reads the actual typefaces and colour values rather than approximating.
What gets cut to save time? Loading screens, cursor travel and half-typed input. Vizard Agent keeps every real action.
Can I get a version with narration too? Yes, from the same cut. Ask Vizard Agent for both.
What if we run over on the day? Ask for a shorter variant in advance. Vizard Agent can drop whole chapters cleanly.
Does the text ever cover the interface? It should not. Vizard Agent keeps descriptions outside the product's own screen.
Can it zoom on small type? Yes, and it costs runtime. Vizard Agent will say how much.
What format for a venue projector? Name it. Vizard Agent exports to the resolution and codec the room needs.
Can it loop? Yes, if you want it running before the talk. Ask for a loopable version.
How do I rehearse against it? Play it once and note where you run out of sentence. Vizard Agent adjusts those holds.