Sentioscope

Speech and voice controls, as each vendor documents them

The background sound that makes a place feel real

Ambience is the general background sound of a location: traffic, rain, the hum of a room. In generated video it arrives with the picture and stays with that clip. Two prompt guides document how to ask for it, and one pipeline documents what happens to it when the dialogue is changed afterwards. As of 2026-10-13.

Four kinds of sound in a generated clipOne prompt guide lists sound effects, ambient noise and dialogue as the three kinds of audio to describe, each in its own sentences. A score is the fourth layer, and in one pipeline it is mixed per episode rather than generated per clip.Front of the mixDialogueSpoken words from a character or a narrator.Sound effectsDistinct sounds inside the scene: a phone, a door, a splash.Ambient noiseThe general background that makes a place feel real.ScoreMusic under the scene, written separately or left out.Underneath
Fig. 1 Ambience is the layer that tells a viewer where the scene is.
Ambience, as published. Recorded 2026-10-13.
VendorWhat the documentation saysWhere it is decided
Google VeoAmbient noise is listed beside sound effects and dialogue, with audio described in separate sentencesIn the prompt
SceneMixerA lip-sync redub does not keep the clip's own ambience; the episode's score is mixed back under the new versionAfter generation
Sora 2Background sound goes in its own block under the shot; a silent shot can carry one small sound as a rhythm cueIn the prompt

Inclusion rule. Vendors whose documentation treats background sound as distinct from dialogue and music, either in how it is asked for or in what happens to it later. A page that lists audio as a feature without addressing ambience is not given a row. Order. Alphabetical by vendor name.

1Asked for in sentences of its own

Veo's prompt guide splits audio into three kinds, sound effects, ambient noise and dialogue, and recommends describing the audio in separate sentences. Sora's guide gives background sound a labelled block of its own under the shot description. Both treat ambience as something written, not something that turns up in the right form by default.

2Natural ambience, and no score

One of Sora's worked examples asks for natural ambience only, faint wind, fabric flutter and street noise, with no added score. That is the instruction most dialogue shots need: the room under the voices, with music placed across a sequence afterwards rather than generated into every clip.

3A silent shot still has a sound

The same guide suggests that a shot with no dialogue can still carry one small sound, a distant traffic hiss or a crisp snap, as a rhythm cue rather than a full soundtrack. In a cut, a clip with no sound at all stands out against the clips either side of it.

4What happens to it when a line changes

Ambience belongs to the take. When SceneMixer redubs a clip by lip sync, the picture is kept frame for frame and every line is voiced again, and the clip's own soundtrack, its ambience and any music the take carried, is not kept; the episode's score is mixed back under the new version on its own. A redub therefore changes the sound of the room as well as the words, which is worth knowing before redubbing a shot set somewhere loud.

5Keeping one place sounding like one place

Because each clip generates its own background, two shots of the same room can sound like two rooms. Writing the same background sound into every shot of a scene, word for word, is the cheapest way to keep a location consistent, for the same reason the same description keeps a face consistent.

6Sources read for this entry

This page defines the term and logs what each vendor documents about it, each figure read on 2026-10-13. Related: Native audio, Reference audio.