How to replace one line of narration in a mix you cannot unmix
Send Vizard Agent the finished mix and a clean recording of the corrected line, and say which phrase is wrong. It locates the phrase by word timing, takes out only that stretch of speech while the music and ambience continue underneath, and drops the new line into the same window.
What do you type into Vizard Agent?
Quote the wrong phrase word for word and say the rest of the mix is untouchable. "Five months later" lets Vizard Agent find the exact word boundaries in a transcript. The negative instructions matter most here, because the obvious shortcut, regenerating the whole audio, destroys exactly what you're trying to keep.
Use the mp3 as the source. Find "five months later" and replace only that phrase with the wav. Keep the original music, effects, ambience and timing exactly as they are.
What does Vizard Agent actually do?
It works out where the phrase is before it touches anything. Word timings give the exact start and end of the spoken phrase, and loudness is measured on both files so the new line is set to match rather than guessed at. Only then does any audio get written. In the session behind this article it:
- Transcribed the full mix and the replacement clip, and read the word-level timing files for the phrase boundaries.
- Measured integrated loudness on both and generated spectrograms to see where the narration band sits against the music bed.
- Found its first attempts at suppressing the old phrase weren't clean enough, so it cut the segment out explicitly and filtered the old speech band instead of turning it down.
- Dropped the replacement line in with short edge ramps so the join wouldn't click, leaving the bed running.
- Rendered at the exact original duration with no codec padding, then transcribed the delivered file to confirm the narrator now says "four".
When does this not work well?
When the replacement line is a different length and the picture is locked. A shorter line leaves a hole; a longer one overruns the next sentence or has to be sped up. Vizard Agent will say which way the difference falls. It also struggles when old speech and music share frequencies at similar levels.
Common questions
Do I have to re-record the whole narration? No. Vizard Agent needs only the corrected line as a separate file.
Will the music dip where the edit is? It should not. Vizard Agent removes only the speech, and the bed runs continuously through the window.
How do I know the old words are gone? Ask Vizard Agent for a transcript of the delivered file. That is a reading of the output, not a report on the process.