How to get rid of the tick you can hear at every cut
Ask Vizard Agent to smooth the edit joins first and then hunt the leftovers separately. A tick created by cutting is fixed with a short crossfade at every join; a tap that was in the original recording survives that entirely, and is removed as a transient after checking on a spectrogram that it is not sitting on top of a word.
What do you type into Vizard Agent?
Describe the sound rather than assuming its cause. "The transitions are glitchy" points at the edit while "there is a tap" points at the recording, and starting from the wrong one costs you a whole round. A timestamp for one tap you can hear is worth a great deal to Vizard Agent.
Prompt
What does Vizard Agent actually do?
It treats the tick as two possible problems and fixes them in two passes. Ticks made by cutting go with short crossfades. Taps that survive were in the recording, and Vizard Agent identifies those on a spectrogram, where a knock against a desk looks nothing like a consonant. In the session behind this article it:
- Checked the joins around the point the user named in a talking-head edit tightened word by word, and measured the audio and video durations so the fix would stay in sync.
- Added short audio crossfades at every edit join, aligned the smoothed audio exactly to the picture and reviewed the waveform for click-free joins.
- Listened around the taps that could still be heard, then built a waveform and a spectrogram of that section to separate handling noise from speech.
- Mapped the tap times against the word timings so nothing would be cut through, working in the gaps between words wherever it could.
- Removed the transients while keeping the speech underneath, then verified that the wording and timing hadn't changed.
When does this not work well?
A tap sitting exactly on a word can't be removed without taking part of the word with it. Very frequent handling noise is a recording problem rather than an editing one, and cleaning it everywhere starts to flatten the voice. Music under the speech masks the taps, and very short gaps between words leave nowhere safe to repair.
Common questions
Why did the crossfades not fix mine? Because those taps were in the recording, which Vizard Agent handles separately.
Will the voice change? Not if the repair stays localised, and Vizard Agent keeps it that way.
Will it go out of sync? Not if the durations are measured first, which is why Vizard Agent does that step.