Capabilities
Language support
Language support for the Inworld first-party STT model.
The Inworld STT model (inworld/inworld-stt-1) supports 30 languages for speech recognition.
Supported languages
| Language | Code |
|---|---|
| Arabic | ar |
| Cantonese | yue |
| Chinese | zh |
| Czech | cs |
| Danish | da |
| Dutch | nl |
| English | en |
| Filipino | fil |
| Finnish | fi |
| French | fr |
| German | de |
| Greek | el |
| Hindi | hi |
| Hungarian | hu |
| Indonesian | id |
| Italian | it |
| Japanese | ja |
| Korean | ko |
| Macedonian | mk |
| Malay | ms |
| Persian | fa |
| Polish | pl |
| Portuguese | pt |
| Romanian | ro |
| Russian | ru |
| Spanish | es |
| Swedish | sv |
| Thai | th |
| Turkish | tr |
| Vietnamese | vi |
Specifying a language
The language field is a language hint: it tells the model which language to prefer, but the model may not follow it. The model detects the spoken language from the audio. You can switch languages during a conversation without changing the hint.
The field accepts ISO 639-1 language codes (e.g., en, ja) matching the codes listed in the table above.
BCP-47 codes (e.g.,
en-US, ja-JP) are also accepted and will be automatically converted to the base ISO 639 language code — for example, en-US becomes en. Regional variants do not affect recognition behavior.If you know the audio's primary language, a hint generally improves transcription accuracy. This is especially useful for short utterances where auto-detection may lack context.
Setting
language also constrains the output script for English, Chinese, Cantonese, Japanese, Korean, Russian, and Hindi. For example, en keeps the transcript in Latin script (a name spoken in another language is romanized rather than written in its native script), while ja allows Japanese script. This applies only when you set the hint: to let auto-detection follow the spoken language as it changes mid-stream, leave language empty.