What MiniMax documents about speech

MiniMax (platform.minimax.io) documents native speech with lip-sync tied to whoever is on screen, and takes reference audio for the voice, capped at 15 seconds across at most three clips. As of 2026-09-12.

Inclusion rule. Fields are the same five for every model here; a field the vendor does not document is recorded as not documented rather than inferred from generated output. Order. Fixed field order, identical on every model page.

1A published figure for how much voice can be supplied

Fifteen seconds in total, across three clips, is the tightest published constraint in this register and the only one stated as a figure rather than a capability. It sets the shape of the work: a voice has to be established from a very short sample, and there is no documented way to store it, so the sample goes in on every call that needs it.

Tying lip-sync to the speaker who is on screen is the other half of the same design. It implies the model is deciding which face moves, which is the part that usually breaks in a two-hander, and it is documented as a behaviour rather than as a control the prompt can override.

2A constraint from elsewhere that lands on dialogue work

A first-frame image and reference images cannot be used in the same call. That restriction is not about audio at all, but it collides with dialogue scenes: continuing from a known frame and carrying a character reference are exactly the two things a conversation between established characters needs at once.

It is recorded here because a register of speech controls that ignored it would describe a workflow the documentation does not permit.

3This entry, one column at a time

Each of these stays inside a single field: what MiniMax puts there, what the wording settles, and what it leaves for a take to answer.

• Audio source — with the picture.
• Voice source — a reference clip on the call.
• Per-character binding — on the call.
• Lip-sync — whoever is on screen.

4Read against another entry

Each of these puts MiniMax beside one other entry on a column where the two land at opposite grades of answer.

• Runway and MiniMax — on voice source.
• MiniMax and Wan 3.0 — on voice source.
• MiniMax and Sora 2 — on lip-sync.

5Where this model sits against the rest

Inclusion rule. Models whose vendor documents something about generated speech. A model with nothing published on the point is still listed, with the cell marked hollow. Order. Alphabetical by model name.

Inclusion rule. Models whose vendor documents something about generated speech. A model with nothing published on the point is still listed, with the cell marked hollow. Order. Alphabetical by model name.

Inclusion rule. Models whose vendor documents something about generated speech. A model with nothing published on the point is still listed, with the cell marked hollow. Order. Alphabetical by model name.

6Sources

Taken from the video generation guide at platform.minimax.io on 2026-09-12. Every entry against the same five fields sits on the speech controls page; the rule behind each field is on how read. The entries either side of this one: PixVerse, Runway.

• Audio source
  Native speech with lip-sync tied to the speaker who is on screenspeaker-awareMiniMax, video generation guide / recorded 2026-09-12
• Voice source
  Reference audio is capped at 15 seconds in total across at most 3 clipsa hard published limitMiniMax, video generation guide / recorded 2026-09-12
• Constraint that collides with voice work
  A first-frame image and reference images cannot be used in the same callMiniMax, video generation guide / recorded 2026-09-12