Entries that publish a limit as a number

Seven entries publish at least one limit as a number rather than as a capability. Seconds, minutes, megabytes and characters. These are the figures a schedule can be built from, and they are the rarest thing in this register. From 7 source pages, read 2026-09-12 to 2026-09-22; no maker pays to be listed.

Inclusion rule. Entries whose documentation states at least one limit as a figure bearing on speech: a duration, a file size, a character count or an allowance. A capability described without a number does not earn a row. Order. Alphabetical by model name.

1A figure is the part of a page a schedule can actually use

Everything else in this register is prose about capability. A figure is different: it can be divided into an episode count, multiplied by a cast size, or compared against a delivery slot. Seven entries publish one, and the other nine leave every quantity to discovery.

Publishing a figure is also a commitment. A vendor that says fifteen seconds has said something that can be wrong, which is more than a bare claim of support does.

2The figures are not measuring the same thing

Ten minutes of output duration, twenty minutes of enrolment audio and fifteen seconds of reference clip look comparable and are not. One bounds a deliverable, one bounds a studio booking, and one bounds how much evidence a model gets about a voice.

Lined up in one table the pattern is clearer: the tightest figures are all about what a production hands over, and the loosest are all about what comes back. Vendors are more willing to bound their inputs than their outputs.

3Two figures repeat, and the repetition is informative

Fifteen seconds in total appears twice, from two different vendors, with different arrangements around it. Fifteen megabytes and ten megabytes sit close enough to suggest a common ceiling for a short sample. A number that recurs across vendors usually reflects a shared constraint rather than a shared decision.

For a production the useful consequence is that fifteen seconds can be treated as the planning unit for reference audio generally. Choose a clean fifteen seconds of ordinary speech per character, store it beside the visual reference, and it will fit wherever a clip is accepted.

4The entries on this route, one page each

Each of these links to that entry on the field this route groups by. The wording behind every cell, and the date it was read, sits on the page it links to.

• D-ID — a voice id, or a recording by url.
• HeyGen — a recording to clone from.
• MiniMax — a reference clip on the call.
• Runway — a 10-second to 5-minute sample, or a sentence.
• Sora 2 — nothing documented as an input.
• sync-3 — nothing documented as an input.
• Wan 3.0 — reference audio, 15 seconds in total.

• 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
• Voice source
  Reference audio is WAV or MP3, one to fifteen seconds a clip, fifteen seconds in total and no more than 15 MBa published capAlibaba Cloud, Wan 3.0 API reference / recorded 2026-09-22
• Voice source
  A custom voice can be built from an audio sample between 10 seconds and 5 minutes long and at most 10 MBa published sample windowRunway, custom voices reference / recorded 2026-09-22
• Shot length a line has to fit
  Generations run 16 or 20 seconds, and an extension adds up to 20 seconds at a time to a total of 120OpenAI, video generation guide / 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
• Audio source
  A script is either text of up to forty thousand characters or an audio url, with audio limited to five minutes for clips and ten for talksa script read aloud or a fileD-ID, create a talk reference / recorded 2026-09-22
• Voice source
  A voice clone is instant from one recording or professional from twenty minutes or more, and is then passed as a voice idtwo enrolment routes with a thresholdHeyGen, API quick start / recorded 2026-09-22

5Speech, generated with the picture

For the speech side of “A figure in writing”, SceneMixer's guides describe how lines are spoken:

AI dubbing replaces the audio of a finished clip and then tries to match the lips. Native-language generation gives the video model the line in the target language before the shot exists, so mouth movement, pauses and shot length are produced for that language.

SceneMixer, Native-Language Dialogue vs. Dubbing in AI Video (2026): Why the Cast Should Speak the Line — “Is native-language generation the same as AI dubbing?” (read 2026-10-13)

SceneMixer writes dialogue into the script analysis, assigns each line to a shot with a speaker and a short delivery note, and the video model performs it — picture and audio generated together, so the mouth forms that sentence in the language you chose.

SceneMixer, AI Dialogue Generator (read 2026-10-13)

6Sources

Membership of this route is decided by the wording in the field it groups by, read from the documentation each vendor publishes. The column itself is on the field note, all five columns are on the speech table, and what counts as documented is on how read. Other routes: A named driver, Named, not explained.