How to split a concert recording into one video per song
Upload the show and tell Vizard Agent you want one complete song per video. Vizard Agent transcribes the stage talk, maps the audio energy across each transition at high resolution, cuts test snippets to verify the boundaries by ear, and renders a separate file for every song.
What do you type into Vizard Agent?
Rule out the alternatives directly and by name. Vizard Agent reads a line like "not shorts, not highlights, one complete song per video" as a hard constraint, and that single sentence changes the whole method from choosing good moments to finding exact boundaries. A setlist helps name the files but is not necessary.
Prompt
What does Vizard Agent actually do?
Finding where a song ends in a live recording is harder than it sounds. The last chord, the applause, the thanks and the count-in for the next number all blur together. An energy curve shows where something changed, but only listening tells you whether that was the song ending or the drummer counting in. In the session behind this article it:
- Probed the recording and transcribed the audio to find the stage talk between songs, then examined the exact word timings around every transition.
- Extracted the audio for acoustic analysis, calculated the energy profile across each transition, and analysed the intros, endings and applause as separate events.
- Extracted frames from each transition and inspected them visually.
- Mapped the sound energy at high resolution across all four transitions, then cut short audio snippets at each candidate boundary and checked them individually.
- Rendered one video per song as three separate files, one of them 1920x1080 at 30fps and 125.53 seconds long, and verified the first and last frames and the audio levels at each entry and exit.
When does this not work well?
A live recording is one continuous performance that happens to contain songs, and not every show divides cleanly. Where a medley or segue runs one song into the next, any cut is arbitrary. Loud crowd noise masks endings, and whether long stage talk belongs to the next song is your call. Splitting a recording also does not change who owns the songs.
Common questions
Why not just cut on the applause? Applause starts before the last chord decays and often runs into the next introduction. Vizard Agent cuts on verified boundaries instead.
What about the talking between songs? Vizard Agent transcribes it, so you can decide whether it belongs to a song or gets cut.
Does this work for other long recordings? Yes. Vizard Agent splits a lecture series, a conference day or a DJ set the same way.