How to use a separate audio track for each podcast guest
Send the camera files and one audio track per guest. Vizard Agent syncs each track to the picture, verifies the alignment holds to the end of the episode, works out who is on which channel, cuts between cameras by who is actually speaking, and gates the bleed out of each microphone.
What do you type into Vizard Agent?
Say what the tracks are. A folder of files doesn't announce that channel three is the guest and channel four is both of them together, and that mapping is what makes the rest of the edit automatic. Without separate tracks, who is speaking is a guess made from watching mouths move, so send every one to Vizard Agent.
Prompt
What does Vizard Agent actually do?
A separate track per person records who was talking at every instant, which is what a multicam edit needs. The cut list is derived rather than judged, and a gate keyed to the channel with the speaker on it can be far more aggressive than one that opens on volume. In the session behind this article it:
- Transcribed the three channels of a two-camera interview, synced the first camera and measured the drift between camera and audio precisely.
- Searched the transcripts for microphone notes and long pauses, and checked whether anyone spoke in pauses the transcript missed.
- Identified who was on each channel, confirmed the guest's name, and verified the sync across the whole episode, not just the start.
- Built the cut list from who was speaking, added listener reaction shots in long stretches, and hid the cut at the microphone note in a camera change.
- Placed the logo and checked it on a real frame.
- Gated the bleed out of each microphone keyed to who was speaking, listened to a test passage for artifacts, and uploaded the cleaned microphone for comparison.
When does this not work well?
Everything depends on the tracks and the picture being alignable, and on each track belonging to exactly one person. Long recordings from separate devices creep apart, a shared microphone gives no per-person information, and heavy crosstalk puts both people on both channels. Very aggressive gating can also clip quiet agreement sounds.
Common questions
Will it know who is who? Vizard Agent works it out from the transcript and the picture, then confirms names.
What if the tracks drift? Vizard Agent measures the drift and corrects it across the episode.
Will gating sound unnatural? Vizard Agent tests a passage and listens for artifacts before applying it.