Resynthesis
Resynthesis cleans audio by rebuilding the speech from a model of what clean speech should sound like, rather than subtracting noise band by band. Because it reconstructs the quiet edges instead of gating them, it tends to preserve word endings better.
Why it matters
Resynthesis is a different answer to the same problem as spectral gating, and it fails in a different way. Instead of subtracting noise and risking the speech, it reconstructs the speech from a model of what clean speech should be. That preserves the fragile edges a gate would mute, so word endings and breaths survive.
The trade is that a model that rebuilds speech can also invent detail, so pushed too hard it introduces artifacts rather than removing them.
In practice
Reach for resynthesis on speech-heavy material where a gate has been eating word endings. Still A and B against the original, and listen for artifacts on held vowels, which is where an over-eager reconstruction shows.
How Vocast handles this
Vocast offers a resynthesis cleanup mode that reconstructs speech rather than gating noise, tuned to keep word endings and breaths rather than filing them off.