What sync-3 documents about speech

sync-3 (sync.so) publishes a language count of ninety-five or more, accepts video or a still image paired with either audio or text, and matches lip movement to whatever track it is handed. As of 2026-09-22.

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.

1The largest language figure here, and still a figure

Ninety-five or more languages, described as the same coverage as the models before it, is the highest number in this register by a wide margin. It is also a count rather than a list, so it answers whether a language is likely to work and not whether a particular one does.

A count that high is plausible for a model working from a waveform, since there is no per-language voice inventory to build. That reasoning is not on the page, so it is not recorded as fact, and a production with one specific market still has to establish coverage the hard way.

2Four accepted pairs, and what each implies about the schedule

Video with audio, video with text, image with audio, image with text. The two audio pairs take a performance that already exists; the two text pairs have the speech produced on the way through. Choosing between them is choosing whether the actors record first.

The image pairs are the interesting ones for drama, because they turn a still into a talking shot without a camera having rolled. A free account runs one generation a month at fifteen seconds, which makes clear that evaluating this properly is a paid exercise.

3Alignment solved, casting elsewhere

Synchronisation is the product, so the lip-sync field fills itself. What stays empty is the voice: nothing on the model page says where one comes from, whether a choice persists, or what happens when two faces are in frame. A model attending to a waveform has no obvious way to know which mouth should move.

For a serial that leaves the same hole as everywhere else in this register, arriving from the opposite direction. The mouth is handled and the cast is somebody else's department.

4This entry, one column at a time

Each of these stays inside a single field: what sync-3 puts there, what the wording settles, and what it leaves for a take to answer.

• Audio source — supplied, or read from text.
• Voice source — nothing documented as an input.
• Languages — a count, ninety-five or more.
• Lip-sync — the audio it is given.

5Read against another entry

Each of these puts sync-3 beside one other entry on a column where the two land at opposite grades of answer.

• Vidu and sync-3 — on audio source.
• Synthesia and sync-3 — on languages.

6Where 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.

7Sources

Taken from the model documentation at sync.so on 2026-09-22. 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: Synthesia, Veo.

• Languages
  sync-3 supports 95 or more languages, described as the same coverage as the models before itpublished as a countSync, sync-3 model documentation / recorded 2026-09-22
• Audio source
  The accepted pairs are video with audio, video with text, image with audio and image with textsupplied or read from textSync, sync-3 model documentation / recorded 2026-09-22
• Length per call on the free tier
  A free account runs one generation a month with a fifteen-second ceiling, with paid limits following the planSync, sync-3 model documentation / recorded 2026-09-22