The finder measures speech density, loudness, changes in intensity and vocal accents, then snaps the in and out points to nearby silences so a cut never splits a word.
Free, no account, no watermark. Everything runs in your browser: no file is uploaded to any server.