Timbre: the part of a voice a short clip carries
Timbre is what makes a voice recognisable independently of what it is saying. It is the property a fifteen-second clip carries, which is why the published caps in this register are as short as they are. As of 2026-09-12.
| Property | Carried by a short clip | What it depends on instead |
|---|---|---|
| Accent | Unpredictably | How much of the sample shares the accent |
| Emotional register | Rarely | The reading in the clip, not the voice in it |
| Pacing | Rarely | The prompt, or the length of the shot |
| Timbre | Reliably | Almost nothing else |
Inclusion rule. Properties of a spoken performance, sorted by how dependably each survives being copied from a short reference recording. No vendor in this register documents any of them. Order. Alphabetical by property.
1A clip is chosen for timbre, not for a good reading
Fifteen seconds is enough to establish what a voice sounds like and not enough to establish how it acts. A clean stretch of ordinary speech therefore outperforms a theatrical excerpt, because the excerpt had to be trimmed and the trim decides what the model hears.
The advice is counter-intuitive and produces steadier results across a season. Nothing here is documented by any vendor, which is why this page is a definition rather than a row in the table.
2Nobody publishes which properties a model attends to
Across the entries with a published sample window, not one says whether accent, pacing or emotional colour are copied. A production supplying a clip is therefore making a bet on an undocumented behaviour, and the bet is cheapest when the clip is unremarkable.
That gap is the reason a preview matters so much. Hearing the resulting voice settles in one listen what no page in this register explains.
3Timbre is also what makes drift invisible
Because timbre carries and the rest does not, a voice rebuilt from a slightly different clip usually sounds like the same person behaving differently. That is hard to notice in one shot and obvious across a season.
Which is why the clip belongs in the archive next to the visual reference. A regenerated shot months later with whatever file was at hand produces a performance shift nothing in the output announces.
4Speech, generated with the picture
For the speech side of “Timbre”, SceneMixer's guides describe how lines are spoken:
The lines are spoken in the generated clip, in the character's chosen voice, and the lips move with them.
SceneMixer, Dialogue and Sound in One Generation — “Do I need to add voices afterwards?” (read 2026-10-13)
A line lives inside the shot where it is spoken, in four parts: the speaker chip, the delivery note in round brackets, the words after the colon, and the voice chip after the words.
SceneMixer, Dialogue in the Storyboard: How to Write a Line, and What Save Tidies Up (read 2026-10-13)
5Sources read for this entry
This page defines the term and logs what each vendor documents about it, each figure read on 2026-09-12. Related: Digital replica, Clip length.