Subtitle: the cheapest language layer, with its own limits
A subtitle is translated text laid over a finished picture. It needs no voice, no model that can pronounce anything, and it can be produced for languages a vendor has no voice for at all. As of 2026-09-22.
| Property | Subtitle | Dub |
|---|---|---|
| Needs a voice | No | Yes |
| Reaches text inside the picture | No | No |
| Constrained by reading speed | Yes | No |
| Constrained by mouth timing | No | Yes |
Inclusion rule. The two ways a finished video reaches another language, compared on what each one needs and what limits it. Order. Fixed order, the cheaper layer first.
1A subtitle count is not a voice count
Because subtitling needs no voice, a subtitle language list can be far longer than a voice list at the same vendor, and the two are not evidence for each other in either direction.
One vendor in this register publishes a subtitle count and no voice list at all. The register records the subtitle figure outside all three columns, because putting it in the voice column would let a reader take it as a claim about speech.
2The limits are published by distributors, not by tool vendors
Characters per line and characters per second are set by the platform a release goes to, and they differ by language and by audience age. A vertical frame tightens the line limit before the reading limit binds.
None of that is in any vendor language list, and all of it constrains what a translation can say. It is the reason a subtitle is often a shorter sentence rather than a faithful one.
3It cannot reach anything drawn in the picture
A sign, a note or a screen generated in one language stays in that language. A subtitle sits on top of the picture and cannot replace part of it.
So the cheap layer has one hard boundary, and it is the same boundary dubbing has. Both stop at the edge of the image, which is why decisions about text in frame belong at the script stage.
A definition, not a measurement: no vendor figure appears here that is not also in the register. The one subtitle figure in this register is recorded outside the three columns.
Nearby terms: In-shot audio, Burned-in text. The whole vocabulary is at terms; nothing on this page is a claim about a named product.