How to keep the recorded sound running when your video cuts to photos

Last updated 2026-09-17 · 8 min read

What is the short version?

You filmed an approach and a doorway, and the reveal itself is a set of photographs. Cut them together and the picture flows while the sound falls off a cliff — footsteps, a door, and then nothing for the last eight seconds.

1. Tell Vizard Agent the recorded sound must run through the stills.
2. Name what you want to still hear: footsteps, the door, the room.
3. Ask it to check the audio stream covers the whole video, not just the fade.

What do you need before you start?

The clips, the stills, and an order for them. Vizard Agent will measure the recorded levels itself, but it needs to know which section is the arrival and which is the reveal, because those two want different amounts of music under them.

• The video clips. With their original sound.
• The stills. In the order they should land.
• What the reveal is. So the pacing builds towards it.
• What must stay audible. Footsteps, a door, voices nearby.
• Whether you want music. And how far under.

What do you type into Vizard Agent?

Ask for the sound explicitly. Almost every request for a cut describes the picture, and an edit built to that description will be perfectly reasonable while going silent over the stills, because nothing in the instruction said the ambience was part of the point.

Prompt

Variants worth knowing:

• "Carry the sound through the photos." The instruction that fixes the cliff.
• "I want the footsteps audible." Names what the mix has to protect.
• "Fade at the end, not at the stills." Where the taper belongs.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a short film built from two phone clips and four stills. The sound work takes more passes than the picture, and a silent tail is found by measurement rather than by listening.

1. Reviews the clips and each still to understand the location and the reveal.
2. Measures the source audio so the merged cut holds one comfortable level.
3. Builds the sequence — exterior approach, entry, interior, then the artwork stills.
4. Checks the still handoffs frame by frame after the lobby shot.
5. Makes the ambience explicit and continuous through the still section.
6. Checks the audio stream's own duration to confirm the fade covers the full video.
7. Generates a music cue and measures its loudness before mixing.
8. Matches the music level to the active ambient section, not to the whole file.
9. Rebalances without automatic attenuation of the original ambience.
10. Replaces an artificial-sounding cue with a warmer acoustic one.
11. Lowers the music and lifts the ambience so footsteps and door sounds stay present.
12. Balances the approach and the reveal separately, then plays back the uploaded copy.

Step six is the check most people never make. A video can look finished, sound finished on a quick scrub, and still carry an audio stream that is two seconds shorter than the picture. Vizard Agent reads the stream duration rather than trusting the fade it just applied.

Step nine is a small decision with a large effect. Automatic attenuation ducks whatever is underneath the music, which in a film like this is the entire recorded world — so the door closing gets quietly pulled down every time the cue swells. Vizard Agent sets the balance by measured level instead.

Step eight matters because the ambience is not constant. Matching music to the average loudness of the whole track makes it too loud during the quiet approach; Vizard Agent matches it to the section where there is actually something happening.

What does the result look like?

One continuous piece. You hear the street, the door, the footsteps in the lobby, and that same room carries under the photographs while the music sits below it and tapers to the end of the last frame. Vizard Agent checks the waveform as well as the picture before it hands the file over.

When does this not work well?

Some recorded sound is not worth carrying. If the clip's audio is wind, handling noise or a conversation that has nothing to do with the film, extending it under the stills spreads the problem rather than solving it, and Vizard Agent will suggest a clean room tone or music instead.

• Wind and handling noise. Continuing it makes it more obvious.
• Speech that ends mid-sentence. A carried tail sounds like a fault.
• Very short clips. Not enough ambience to stretch over the stills.
• Heavily compressed source audio. It pumps when levelled.
• A long still section. Minutes of room tone reads as a mistake.

How do you fix a result that came back wrong?

Say what you can and cannot hear, and at which point in the video you heard it. Vizard Agent treats the approach and the reveal as two separate balances, so a note about one of them can be acted on without disturbing the other, and a single sentence is usually enough to place the fault.

• "It goes silent on the photos." The ambience is extended and made explicit.
• "The music drowns the footsteps." The cue comes down, the ambience goes up.
• "The music sounds fake." A different cue is generated and measured again.
• "It stops instead of ending." The fade is set against the stream duration.

How does Vizard Agent compare to doing it yourself?

By hand the silent tail is the classic mistake, because you build the picture first and the stills genuinely have nothing to play. You notice on the third viewing, when the film is otherwise done, and fixing it means going back into the mix and rebalancing everything you had already settled.

Common questions

Why does the sound stop at the photos? Because a still has no audio track. Vizard Agent carries the recorded ambience across it.

Can it use a different room tone? Yes. If the recorded sound is unusable, Vizard Agent will suggest a clean bed instead.

Do I have to add music? No. Ambience alone works, and Vizard Agent will mix it to one consistent level.

Why not let the music duck automatically? Because it ducks the footsteps too. Vizard Agent sets the balance by measurement.

Can I choose the style of music? Yes, and changing it is cheap — Vizard Agent measures each cue before mixing.

How does it know where the reveal is? From your description of the story, which is why naming it is worth a sentence.

Will the stills be too long on screen? Tell Vizard Agent how long you want each one, or let it pace them to the music.

Can it mix video clips with different levels? Yes. Vizard Agent measures each source before merging them.

What if the video should be vertical? Say so. Vizard Agent rebuilds the master in that frame rather than cropping the export.

Does a title card interrupt the sound? No. The sound runs underneath whatever the picture is doing.

How do I check it myself? Play the last ten seconds with your eyes shut — that is where the fault lives.

Can it fade at the end instead? Yes, and that is usually right. Vizard Agent tapers it across the final frames.

What about a voiceover on top? It sits above both, and Vizard Agent balances the ambience under it.

Will the uploaded copy sound the same? Vizard Agent plays back the uploaded file specifically to confirm that it does.