How to check which music actually ended up in your video

Last updated 2026-09-05 · 8 min read

What is the short version?

Arguments about the music are usually not arguments about the music. They are arguments about which thirty seconds of a file somebody meant, and they end the moment the bed that was used can be listened to on its own.

1. Go to Vizard Agent and say the music is wrong.
2. Ask for the bed used in the render, isolated.
3. Say which part of your file you actually meant.

What do you need before you start?

Only your original file and a willingness to be specific about which part of it you meant. That second part is where these disagreements are settled, because "the music I sent" is a complete description of a file and an incomplete description of a piece of music inside it.

• The file you uploaded. The exact one.
• Which section you meant. The opening, the strings, the melody.
• Where in the video it should sit. Under the whole thing or one part.
• How loud. Behind the voice, or leading.
• Whether anything else may change. Usually nothing.

What do you type into Vizard Agent?

Say it plainly and ask for the evidence rather than a rebuild. Asking to hear what is actually in the render is faster than another round of guessing, and it is the step that tells you whether the fault is in the file, the section or the mix.

Prompt

Variants worth knowing:

• "Let me hear the bed on its own." Ends the disagreement.
• "What is in my file?" Often the real answer.
• "No other music." Rules out generated beds entirely.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real narrated film where the client was certain the wrong track had been used. Notice that nothing is rebuilt until both sides know what is actually in the render and what is actually in the upload.

1. Verifies which file it downloaded from the upload.
2. Proves which track is in the render rather than asserting it.
3. Makes an isolated copy of the bed used so it can be listened to alone.
4. Checks the newly submitted track against the earlier one.
5. Compares the two uploads and samples the new file.
6. Listens through the piece section by section.
7. Examines the string theme and then the melody section separately.
8. Shapes a new bed from the section you meant.
9. Calibrates the bed level against the narration.
10. Maps the whole submitted file from beginning to end.
11. Splits the compilation into its six separate tracks.
12. Identifies each of the six so the right one can be named.
13. Renders both cuts with the section you meant underneath.

Steps ten to twelve are the resolution, and they are worth expecting. A file described as "my piece of music" was a compilation, so an instruction to use its first thirty seconds pointed at a completely different piece from the one the client had in their head — and neither side was wrong about anything except which seconds they were discussing.

Step three is the step that shortens all of this. Hearing the bed on its own, without narration over it, converts an argument about impressions into a question with an answer, usually within a minute of listening.

What does the result look like?

Either the music is confirmed correct and you stop worrying about it, or the actual mismatch is identified — a different section, a different piece inside the same file — and the bed is rebuilt from the part you meant, at a level set against the narration.

Both outcomes are better than another round of swapping files.

When does this not work well?

This settles which audio is present, not whether it is licensed or whether it suits the film. It also cannot help when the file supplied is genuinely not the piece you have in mind, which happens more often than anyone expects.

• The wrong file was uploaded. No inspection finds a piece that is not there.
• A streaming link, not a file. Supply audio you can send.
• Licensing questions. Identification is not permission.
• Taste disagreements. "Wrong feel" is a different conversation.
• Heavily processed beds. A ducked, filtered bed sounds unlike the original.

How do you fix a result that came back wrong?

Name the section rather than the file. Vizard Agent keeps the map of your upload and the isolated bed, so "use the strings, from about two minutes in" is a precise instruction where "use my music" was not.

• "Use the melody section." The bed rebuilt from those bars.
• "It is too quiet." The level recalibrated against the voice.
• "That is still not it." The upload re-mapped and its parts listed.
• "Start it later." The bed re-shaped to enter where you say.

How does Vizard Agent compare to doing it yourself?

By hand this becomes several rounds of re-uploading the same file and re-rendering, because nobody has heard the bed on its own and nobody has looked at what the file contains. The compilation problem in particular is invisible until somebody maps the upload.

Common questions

Can it tell me what is in my own file? Yes. Vizard Agent maps it and lists what it finds.

Can I hear the bed without the voice? Yes. Vizard Agent exports an isolated copy for you.

What if my file is a compilation? Vizard Agent splits it and tells you where each piece starts.

Can I ask for one specific section? Yes, and that is the instruction Vizard Agent can act on precisely.

Will it use generated music instead? Not if you say no other music. Vizard Agent uses only your file.

Does the video get re-rendered? Only once the music is settled. Vizard Agent checks first.

Can it match the bed to the narration? Yes. Vizard Agent measures both and sets the balance.

What if I want it to build? Say so. Vizard Agent shapes the bed's level over time.

Can it identify a track I do not know? Vizard Agent can describe it, but identification is not the same as licensing.

Will it loop a short piece? Yes, if you ask. Vizard Agent joins it on a musical boundary.

Can it prove which file it downloaded? Yes. That is the first thing Vizard Agent checks.

What if I change my mind afterwards? Say so. Vizard Agent rebuilds from the section you name.

Does this apply to voiceover files too? Yes. The same inspection works on any supplied audio.

Why does this happen at all? Because "the first thirty seconds of my music" is precise about a file and vague about a piece, and long uploads often contain several.