Sora 2: dialogue guidance that never names a language
This is the one entry here that tells a writer where dialogue belongs in a prompt, how to label speakers and roughly how much speech fits a shot. Across all of that, the language the line is spoken in is never raised. As of 2026-09-22.
| What the documentation settles | What it leaves to a take |
|---|---|
| Dialogue belongs in a dialogue block below the prose description | Whether that block behaves the same in a language other than English |
| Speakers should be labelled consistently, with turns alternated | Whether labels in another script are read as reliably |
| Neither a list nor a count of languages appears in the guide | Which languages the model will actually perform |
Inclusion rule. Only statements found on the vendor page named in the caption, and only ones bearing on this one field. Nothing here is filled in from generated output. Order. In the order the statements arrive on the vendor's page.
1The most detailed dialogue guidance in the register, with a hole in it
Everything a writer needs to format a spoken exchange is on these pages except the language it will be spoken in. That combination is genuinely strange: the vendor has thought about two-handers, about pacing against clip length, and about how a line is marked up, and has not addressed the first question a non-English production asks.
The likely explanation is an English-first document rather than an English-only model. This register does not fill that in, because assuming a default language is exactly the kind of inference that turns into a commissioning mistake.
2Labels and language are the same problem one level up
A speaker label routes a line to a face inside one generation. A language decides whether the line exists at all in a given market. The guide solves the first carefully and leaves the second unmentioned, which means a production can write a technically perfect prompt for a market it cannot verify.
The practical move is to treat language as a test rather than a specification here: generate a short exchange, listen, and keep the result beside the shot list. That is a finding about one clip, which is why it does not belong in this column.
3The other entries that never say which language
Eight more entries reach this column empty. Most of them say little about dialogue at all, which makes this one the outlier: detailed on how to write a line, silent on what language it is in.
- LTX Studio — not documented by the vendor.
- Luma Ray — not documented by the vendor.
- MiniMax — not documented by the vendor.
- Runway — not documented by the vendor.
- Veo — not documented by the vendor.
- Vidu — not documented by the vendor.
- Wan 3.0 — not documented by the vendor.
- Wan2.2-S2V — not documented by the vendor.
- LanguagesNot documented by the vendor (as of 2026-09-22)neither a list nor a count
- How a spoken line is written into a promptDialogue belongs in a dialogue block below the prose description, with exchanges limited to a handful of sentences
- Two characters in frameFor multi-character scenes the guide asks for speakers labelled consistently and turns alternated, so each line lands on the right characteraddressed rather than skipped
4Sources
Read from the model page and prompting guide at developers.openai.com on 2026-09-22. The same column across every entry is on languages; everything this vendor publishes about speech is on Sora 2. What counts as documented is on how read.