Separating a picture revision from an audio one
Shots get redone for visual reasons constantly. Whether that costs a performance depends entirely on whether the audio came from the same call, and most entries in this register give no way to separate the two. As of 2026-09-12.
| Architecture | A picture fix costs | A line fix costs |
|---|---|---|
| A separate speech call | The render only | One cheap speech call |
| Sound generated with the picture | A new performance as well | A generation |
| Sound supplied as a recording | The render only | A room and a performer |
Inclusion rule. The three documented arrangements, compared by what a revision costs on each. Order. Alphabetical by architecture.
1The expense nobody puts in a plan
An approved line vanishing because a background was wrong appears on no rate card. The replacement file is valid, aligned and different, so there is nothing to diagnose and nothing to appeal to.
Productions that have been through it approve picture and audio in one pass rather than two, which is slower and avoids re-opening a decision that was already made.
2A separate speech stage survives revision most easily
Where speech comes from its own call, the audio is an artefact with a name. A picture fix leaves it alone, a word fix costs one cheap call, and the two never collide. That is the arrangement to prefer if a script is still moving.
What it does not give is a performance. A synthesised read can be regenerated endlessly and will not become a different interpretation, because nothing in these products takes direction.
3Supplied audio trades one cost for a bigger, more predictable one
Handing over a recording makes visual revision free and line revision expensive. For a series with a locked script and restless art direction that is the right way round; for one still finding its dialogue it is exactly wrong.
The choice is worth making deliberately at the start, because it is effectively irreversible once a season is in production. Changing architecture mid-run means re-establishing every voice.
4Where the published detail is logged
What is described above is how the techniques work in general. Which models document which of them, in whose words, is kept in the speech controls table with a date on every field.
- Regenerating a shot — what happens to the voice
- Guide track — audio in sync but not for delivery
A mechanism note rather than a documented field. Nothing here is attributed to a model, and nothing here is a claim about one. The sourced material is on the speech controls table. Related: Framing and the mouth, Auditioning first.
5How one tool handles speech
For the speech side of “Two kinds of fix”, SceneMixer's guides describe how lines are spoken:
A SceneMixer segment comes out with its sound already in it: each line spoken in that character’s voice, the speaker’s lips moving with the words, and the ambience of the place — rain, traffic, a clock.
SceneMixer, Dialogue and Sound in One Generation (read 2026-10-13)
Dubbing replaces the audio of a finished clip; the picture was generated for a different line, so mouth shapes, pauses and shot length all belong to the original language. Native-language generation gives the video model the line in the target language before it renders the shot, so the performance, the pacing and the cut points are built for that language.
SceneMixer, Native-Language Dialogue vs. Dubbing in AI Video (2026): Why the Cast Should Speak the Line (read 2026-10-13)