PixVerse: an uploaded recording or a script, never both

A lip-sync request carries its audio as the id of an uploaded recording, or as a speaker id with the text to read. The speaker list holds built-in and custom voices, and the pages give no duration or size limit for the call. From 3 source pages, read 2026-10-25; no maker pays to be listed.

Fig. 1 Cutting scenes at a line break, rather than at a limit nobody published, survives a later change to the endpoint.

PixVerse on audio source, statement by statement. Read from the vendor's lip-sync reference and voice pages on 2026-10-25.
What the documentation settles	What it leaves to a take
Audio is either an uploaded file's id or both text-to-speech fields, and the two modes are not mixed	Whether one changed word can be patched without redoing the whole line
The speaker list returns built-in voices and the account's custom voices	How a custom voice built from a recording compares with a built-in one
Uploads accept MP4, MOV or WebM video and MP3, WAV, M4A or AAC audio; each endpoint sets its own limits	Where the duration and size limits for lip sync actually sit

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.

1Two modes and a rule against blending them

The page states the rule in its opening sentence: for audio, choose a recording or the text fields, and do not mix them. That keeps the request simple and rules out the middle case a production sometimes wants, a recording carrying the performance with a synthetic read patched in over one changed word.

A changed line therefore means a new request, with the whole line either recorded again and uploaded or read again by a voice.

2Limits that are acknowledged and not stated

The upload page lists the accepted containers and says each downstream endpoint imposes its own duration, codec, dimension and file-size limits. The lip-sync page adds that an accepted upload does not guarantee that the media meets the operation's processing requirements.

The useful habit is to run the longest take a scene will need through the call before a schedule depends on it, and to cut at a line break so that scenes survive a later change to a limit nobody published.

3The others that synthesise before they render

Three more entries make speech in a stage of its own. They differ on whether that stage has published limits, and on whether a voice built there can be reused.

D-ID — from a script, or a supplied url.
HeyGen — from a speech endpoint, then rendered.
Synthesia — from the script, or uploaded.

Audio source
Audio on a lip-sync request is either the id of an uploaded audio file or both text-to-speech fields, a speaker id and the text, and the two modes are not mixeda supplied recording or text read by a voicePixVerse, Lipsync reference / recorded 2026-10-25
Voice source
The speaker list returns the built-in voices and the account's custom voices, and can be filtered to either kindcatalogue or supplied samplePixVerse, Get TTS Speaker List reference / recorded 2026-10-25
Containers accepted
Uploads accept MP4, MOV or WebM video and MP3, WAV, M4A or AAC audio, and each downstream endpoint sets its own duration, codec, dimension and file-size limitslimits left to each endpointPixVerse, Upload Media reference / recorded 2026-10-25

4How one tool handles speech

For the speech side of “PixVerse, audio source”, SceneMixer's guides describe how lines are spoken:

The score is written for the episode and mixed to sit under the voices.
SceneMixer, Background Music in an AI Short Drama: The Built-in Score — “Can I choose the track, or change its volume?” (read 2026-10-13)

Voices are matched from a library of sampled preset voices, or supplied by you as an upload of your own audio.
SceneMixer, AI Character Voices — “Can SceneMixer clone a voice?” (read 2026-10-13)

5Sources

Read from the lip-sync reference and voice pages at docs.platform.pixverse.ai on 2026-10-25. The same column across every entry is on audio source; everything this vendor publishes about speech is on PixVerse. What counts as documented is on how read.