Supported languages and recognition models
A recognition model is a model trained to recognize speech in a particular language. The models are trained on datasets generated by Yandex services and applications. This allows us to continually improve speech recognition quality.
The main supported model for each recognition type is the general model. It recognizes speech on any topic in a given language, including short and long utterances, names, addresses, dates, and numbers.
Version tags
Three versions of the general model can be available at the same time. You can select the one you need by tags:
general: The main version of the model.general:rc: The release candidate version available for testing.general:deprecated: The previous version of the model.
Note
We stop supporting the general:deprecated tag versions as new models are released: SpeechKit guarantees two weeks of support for the previous version after we update the general tag version. You can find the list of updates in Yandex SpeechKit release notes: Speech recognition.
You can also use the deferred-general tag for asynchronous recognition with the API v2. Learn more about asynchronous recognition modes.
Supported recognition languages
Specify the language in ISO 639-1 format. All available language code values are case insensitive.
| Code | Language |
|---|---|
auto |
Automatic language recognition |
de-DE |
German |
en-US |
English |
es-ES |
Spanish |
fi-FI |
Finnish |
fr-FR |
French |
he-HE |
Hebrew |
it-IT |
Italian |
kk-KZ |
Kazakh |
nl-NL |
Dutch |
pl-PL |
Polish |
pt-PT |
Portuguese |
pt-BR |
Brazilian Portuguese |
ru-RU |
Russian (default) |
sv-SE |
Swedish |
tr-TR |
Turkish |
uz-UZ |
Uzbek (Latin script) |
Recognition accuracy
In Yandex DataSphere, you can assess the recognition quality of a SpeechKit model yourself using your data.