Which containers and file sizes are accepted
A format and a size ceiling remove a whole class of failed call. Four entries publish them: three name the containers they accept, and one caps a voice sample at ten megabytes. From 4 source pages, read 2026-09-12 to 2026-10-25; no maker pays to be listed.
| Model | What is documented | Detail |
|---|---|---|
| MiniMax | WAV or MP3 | Fifteen seconds in total, three clips |
| PixVerse | MP3, WAV, M4A or AAC audio; MP4, MOV or WebM video | Each endpoint sets its own size and length limits |
| Runway | At most ten megabytes | A sample of ten seconds to five minutes |
| Wan 3.0 | WAV or MP3, up to fifteen megabytes | One to fifteen seconds a clip |
Inclusion rule. Entries whose documentation names an accepted container or a maximum file size for supplied audio. An entry accepting audio without publishing either does not earn a row. Order. Alphabetical by model name.
1The least glamorous documentation is the most immediately useful
Nobody chooses a platform for its container support, and a team that does not know it loses a morning to rejections. Publishing WAV or MP3 with a byte ceiling settles what a production has to prepare before it writes any code.
The numbers are also small enough to be reassuring. Fifteen megabytes will hold far more than fifteen seconds of speech at any sensible bitrate, so the duration is the binding constraint and the size limit is a guard rather than a design consideration.
2A size ceiling and a duration ceiling can bind in either order
One entry here names the containers it accepts and then says each downstream endpoint sets its own duration and size limits, without printing them. That leaves a production knowing a ceiling exists and not where it sits.
Treating the duration as the unit of work rather than as an upper bound avoids the problem. Scenes cut at a line break instead of at a limit survive a later change to a figure nobody published.
3Twelve entries publish neither
Most vendors here accept audio somewhere and never say in what form. A team can usually work it out from an error message, which means the cost of the omission is a day rather than a decision.
It still belongs in the record. A vendor willing to publish a container name has documented its endpoint rather than its model, and across this register that correlates with publishing everything else.
- Voice sourceReference audio is capped at 15 seconds in total across at most 3 clipsa hard published limit
- Voice sourceReference audio is WAV or MP3, one to fifteen seconds a clip, fifteen seconds in total and no more than 15 MBa published cap
- Voice sourceA custom voice can be built from an audio sample between 10 seconds and 5 minutes long and at most 10 MBa published sample window
- Containers acceptedUploads 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 endpoint
4How one tool handles speech
For the speech side of “Formats and file sizes”, SceneMixer's guides describe how lines are spoken:
Scene order, dialogue and beats follow the script; what the pipeline adds is coverage — which shot sizes, who is in frame, how long each shot holds.
SceneMixer, Script to Video, End to End — “How faithful is the shot list to my script?” (read 2026-10-13)
The lines are spoken in the generated clip, in the character's chosen voice, and the lips move with them.
SceneMixer, Dialogue and Sound in One Generation — “Do I need to add voices afterwards?” (read 2026-10-13)
5Sources
Every line is read from the vendor documentation linked on the speech controls page, checked 2026-09-12. What counts as documented is on how read. Nearby: What sets the length, Charging for audio, Asking for an accent.