Dialogue
Also known as: speech, spoken audio
Spoken words in a recording, which behave differently from every other kind of audio because losing a syllable loses the meaning.
Updated
What it means
Dialogue is any spoken content on the track, whether a piece to camera, a comment between riders or something caught behind the Ambient sound of the location. The defining property is that it carries information.
That changes how it can be edited. Music can be cut anywhere a beat falls and ambient sound almost anywhere at all, while speech can only be cut between phrases without damage.
It also outranks the rest of the mix. If a viewer can hear speech at all they will try to follow it, so anything competing with it has to step back, which is what Audio ducking exists for.
How it works in practice
Edit points go in the gaps. Detecting where talking starts and stops is what lets an automatic cutter avoid slicing a word, and Speech-to-text supplies those boundaries as a side effect.
Levels matter more than for anything else on the track. Speech two decibels too quiet is simply not followed, and the viewer stops attending rather than leaning in.
On action footage the realistic goal is often to use it sparingly. A single clear line at the start of a video is worth more than ten half audible ones spread through it.
What to watch out for
Wind destroys speech more thoroughly than it destroys anything else. A rumble that merely spoils music makes words unrecoverable, and no amount of filtering brings a buried voice back.
Cutting mid sentence is the most noticeable editing error there is. A viewer forgives a missed beat instantly and remembers a chopped word for the rest of the video.
Captioning changes the stakes too. Once words are on screen through Subtitles and captions, a mistranscribed line is visibly wrong rather than merely unclear, and a viewer who spots one stops trusting the rest. Getting the words right therefore matters more than getting them on screen quickly.
How RawClip handles it
Where speech is present, cut points are placed between phrases rather than through them, and the music steps back underneath so the words stay intelligible. Camera audio that is mostly wind is left under the music instead of being pushed forward.