Whether a voice can be heard before a shot exists
Casting is normally a listening decision. Two entries in this register document something to listen to before a shot is made: a voice that reaches a ready state and arrives with a preview, and a library in which every preset has a sample to play before it is attached. From 5 source pages, read 2026-09-22 to 2026-10-13; no maker pays to be listed.
| Model | What is documented | Detail |
|---|---|---|
| HeyGen | A clone, enrolled and then used by id | No preview step documented |
| Runway | A ready state with a preview | Before any shot is generated |
| SceneMixer | A sample for every preset, played in the picker | Before the voice is attached to a character |
| Synthesia | A catalogue with names, codes and ids | No sample documented on the page read |
| Wan 3.0 | Reference audio supplied on the call | The result is heard with the shot |
Inclusion rule. Entries whose documentation describes a step that returns audio for approval before a generation. A catalogue that can be browsed without a documented sample does not earn a row. Order. Alphabetical by model name.
1A preview is the difference between choosing and discovering
Where a voice can be heard first, casting is a decision somebody takes deliberately. Where it cannot, the first generation of each character is the audition, and whatever comes back is what the series has, because nothing describes a way to ask for a second opinion.
Two arrangements make that possible. An asynchronous create puts an object between the request and the shot, so there is something to play. A fixed library with a pre-rendered sample for each voice does the same from the other end: the voice exists before anyone asks for it, and changing the choice costs nothing.
2Browsing a catalogue is not the same as hearing one
A published catalogue with language codes and ids is the strongest inventory in this register and it is a list rather than a set of samples. A reader can shortlist by market and gender, and still cannot tell which of two voices suits a dramatic read.
Most platforms with catalogues almost certainly have samples somewhere in a user interface. This register records the documentation a production reads before committing, and a sample player that is not documented cannot be planned around.
3Everywhere else, the shot is the audition
Eleven entries publish nothing on this at all. For the ones that take a reference clip, the nearest thing to an audition is generating a short shot and listening to it, which is a paid test rather than a documented step.
That is worth budgeting for explicitly. One short generation per character before a season starts is cheap, and it is the only way most of this register answers a question that casting cannot skip.
- Per-character bindingA voice is created asynchronously, reaches a ready state with a preview, and is then referred to by its own ida stored voice object
- 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
- Voice sourceA 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 threshold
- LanguagesEach voice is listed with its formal and native language name, a language code, a gender, a name and a voice idpublished as a catalogue with codes
- 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
- Hearing a voice before using itEvery preset in the library has a sample to play in the picker, and matching or changing a match costs nothing
4Sources
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: Outside suppliers named, The column always filled, Speech without the bed.