Voice mode uses your mic — allow permission when asked. Works in Chrome, Edge, and Android Chrome. On iPhone, Safari only (and if voice acts up there, Auto mode always works).
The amber waveform is your voice — silence shows flat. Click the waveform to jump, scrub to a spot, then snap Start here / End here, or drag the sliders. Export replays your selection and re-captures it, so a 60-second trim takes about 60 seconds — the trimmed take then appears in your Takes panel.