Audio

Video tracks play their clips' sound, and audio tracks hold music, voiceover and effects. The Audio workspace (⌥2) docks the mixer beside the viewer and makes audio tracks tall, with waveforms and volume lines.

Clip volume and fades

The inspector's Audio section sets a clip's volume (0 to 200%) and its fade in and fade out.
Clips with sound draw their waveform, and a volume line you can drag up or down to set the level, as in Premiere.
Volume can be keyframed like position and scale (see keyframes); the volume line follows the keyframes.
Right-click a clip to mute or unmute it, or to detach a video clip's sound onto its own audio track.

The mixer

Open the Mixer panel in the sidebar, or switch to the Audio workspace to have it docked beside the viewer. There is one strip for every track that can make sound (audio tracks, and video tracks with sound on them):

Level: the track's volume, 0 to 200%.
Pan: left to right.
Mute and Solo: when any track is soloed, only soloed tracks are heard.
A live peak meter from -48 dBFS to 0, green to red.
Tone: a three-band EQ (Low around 120 Hz, Mid around 1.2 kHz, High around 8 kHz, each ±12 dB; double-click a band to reset it) and a one-knob Compressor. Presets: Voice, and Music bed. They are heard as you play and applied the same way on export.

The Master strip at the bottom shows the left and right meters of the whole mix. Measure reads the loudness of the mix as it would export (LUFS and true peak), and Auto-mix sets each track's level for you: speech around -16 LUFS, music about 8 dB under it with ducking on, and the Voice preset on the voiceover track. It is one step you can undo. Agents use auto_mix and measure_loudness.

Ducking under the voiceover

Mark the track with your narration as the voiceover track: track menu → Voiceover.
On the music track, choose Duck under voiceover ("Lower while the voiceover speaks").

The music dips while the voiceover speaks, in playback and in the export. The track header shows "Ducks under voice".

Voiceover

Cue has a small recording booth for narration. The script is a list of timed lines on the timeline; you record takes for each line against a teleprompter and keep the best one. The Voiceover workspace (⌥4) puts the script beside the viewer with the prompter on the picture.

The script

Open the Voiceover panel. Add lines with New line or Add line at playhead, or import a script from an .srt or .json file.
Each line has where it Starts, a Target length and a Max length. Its status shows whether its take fits: Fits, Tight, Too long or No take.
The line menu can play the line, import a take from a file, rewrite the line (Fit, Shorter or Clearer, with the text model) or delete it. Deleting a line keeps its takes in the Media panel.
Script from footage in the Generate panel turns the speech in a video into script lines, timed where they are spoken, so you can re-record it.

Recording takes

Choose your microphone in the Voiceover panel.
Select a line and press R (or Record). Playback rolls from the pre-roll and the line appears on the teleprompter.
Press Space to stop and keep the take, or Esc to discard it. Recording can also stop by itself after the line's max length.
Every take is kept. Listen to them and choose Use this take for the best one.

Pre-roll, post-roll, trim padding, the silence level, auto-stop and what you hear while recording (nothing, or the timeline) are in the project settings under Recording.

Generated voices

Generate a voice take on a line speaks it with a voice instead of recording, and the Generate panel can voice every missing line at once. Use a macOS voice on your Mac, or OpenAI's voices with delivery instructions ("calm, warm, unhurried"). A generated take is a take like any other.

Denoise and loudness

Reduce background noise in the clip's Audio section has three settings. Light is a high-pass filter and FFT noise reduction for low rumble and steady hiss. Voice (ML) runs RNNoise, a small neural network trained on speech, on your Mac: it removes most sound that isn't a voice (fans, traffic, keyboards). The preview plays the cleaned sound, so what you hear is what you export. You can also ask an agent ("Denoise the interview clips with the voice model").
Normalise loudness in the project settings (Export section) evens out the level of the exported mix.

Beats and cutting to the music

In the Generate panel, the Music section works on a music item from your project:

Button	What it does
Mark the beats	Finds the tempo (BPM) and puts a green Beat marker on every beat where the music is used on the timeline. It warns when the pulse is weak.
Cut to the beat	Moves each cut on a video track to the nearest beat with rolling edits, so the timing of everything else stays.

Clear the beat markers from the markers list (Clear beats) when you are done.

Agents can mark only every Nth beat and set how far a cut may move with detect_beats and snap_cuts_to_beats. To cut pauses out of speech, ask for remove_silence.