What Hedra documents about speech

Hedra (hedra.com) inverts the usual arrangement: audio is a required input, the character in the image lip-syncs and moves to whatever is supplied, and the length of the take sets the length of the shot, up to ten minutes. 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.

1When the audio is the input, alignment stops being a promise

The guide states that avatar videos are driven by audio and that the character in the image will lip-sync and move to the audio provided. Synchronisation is not a feature the model might achieve; it is the task it was handed. That makes this the plainest lip-sync statement in the register, and it is plain because of the architecture rather than the copywriting.

The trade is that a performance has to exist first. A production on this route is casting and recording before it generates anything, which is an older order of work than prompting a scene and hearing what comes back.

2Length decided by the take, not by a parameter

The audio normally determines the video length, and omitting the duration follows the source audio. For a register that spends most of its time on ceilings measured in single-figure seconds, a ten-minute maximum is a different world, and a line that runs long produces a longer shot instead of a rushed read.

It moves the timing problem back into writing and recording, where a director can hear it. Nothing has to be trimmed to fit a four-second window, because the window follows the words.

3Voices are available, and the language claim carries no list

Speech can be generated inline by naming a voice id from the voices endpoint, so a production need not arrive with a recording at all. What the model page then claims is full multi-language support without naming a single language, which is the count-free version of a gap several entries here share.

Nothing is published about holding a voice to a character across shots. With audio supplied per call, consistency is whatever a production's own choices make it, which is workable and unstated.

4This entry, one column at a time

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

• Audio source — supplied; audio is required.
• Voice source — a track, or a voice id from the endpoint.
• Per-character binding — not documented by the vendor.
• Languages — a claim with no names.
• Lip-sync — the supplied audio.

5Read against another entry

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

• Veo and Hedra — on audio source.
• D-ID and Hedra — on audio source.
• Hedra and Synthesia — on lip-sync.

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 Character 3 model page and avatar video guide at hedra.com 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: HeyGen, Kling AI.

• Audio source
  Avatar videos are driven by audio, and the character in the image will lip-sync and move to the audio providedhanded to the model rather than made by itHedra, avatar video guide / recorded 2026-09-22
• Shot length set by the take
  The audio normally determines the video length, and omitting the duration follows the source audioHedra, avatar video guide / recorded 2026-09-22
• Voice source
  Speech can be generated inline instead of uploaded, by naming a voice id drawn from the voices endpointa catalogue behind an endpointHedra, avatar video guide / recorded 2026-09-22
• Languages
  Described as text, image and audio to video with full multi-language support, and a maximum duration of ten minutesa claim carrying no listHedra, Character 3 model page / recorded 2026-09-22