Who separates what it can hear from what it can speak
One vendor in this register publishes two lists: 72 languages it can detect in an uploaded file and 29 it can dub into. Every other spoken-language figure here is one-sided, and none of them says which side it describes. As of 2026-09-12.
| Tool | What is published | Detail |
|---|---|---|
| VEED | 72 detectable in the source, 29 it can dub into | The only two-sided list in the register |
| HeyGen | Two lists, split by purpose rather than direction | Voices against translation targets |
| Everyone else | One figure or one list, direction unstated | A reader cannot tell which side is described |
Inclusion rule. Vendors publishing more than one spoken-language list, plus the rest of the field as one row. Direction means input against output, not purpose. Order. The two-sided list first, then the split-by-purpose list, then the field.
1Forty-three languages it can hear and not speak
Recognising speech and producing it are separate systems with separate costs. Recognition trains on recordings of people talking, of which a great deal exists; synthesis needs a voice built and licensed for each language.
So the asymmetry is structural and every dubbing vendor has it. This is the only one whose pages let a reader see it, which makes the smaller figure the specification and the larger one the intake.
2A one-sided figure with no direction is close to unusable
A vendor publishing a single count has stated neither the unit nor the direction. A reader has a number that might describe what the system can read, what it can produce, or both, and no page says.
Where a vendor publishes only a marketing figure, the safest assumption is that it describes the more flattering side. The register does not make that assumption; it records the wording and notes that the direction is unstated.
3Publishing both sides costs a vendor the bigger number
The smaller figure is the one a buyer purchases, and printing both means printing the smaller one. That is a disclosure decision rather than a product property, and one vendor here has made it.
It is also the cheapest improvement available to everyone else in this column. Four vendors publish a figure and no list at all, and three publish lists without saying which direction they apply to.
4The three columns behind this question
| Column | How many of the 18 tools name a list |
|---|---|
| Dialogue language | 5 of 18 tools |
| Document language | 1 of 18 tools |
| Site language | 9 of 18 tools |
Inclusion rule. The three columns this register keeps. A column stays listed even when no vendor names a list, because silence in a column is itself recorded. Order. Alphabetical by column name.
- Voice languagesA list headed “Supported spoken languages:” with 72 entries, introduced as “the languages that we can detect in the original file”input side of the dubbing list
- Voice languagesA second list headed “Supported Target languages” with 29 entries, introduced as “the languages we can dub to”output side of the dubbing list
- Voice languagesNeither list carries a total; the two are printed one after the other on the same pagehow the lists are presented
5Sources
Counts and lists are read from the vendor pages named in the register, checked 2026-09-12. The counting rule is on how counted. Nearby: Limits per job, Pages on other hosts.