Models & languages
This page is the single reference for what each capability does and which languages it supports today.
Speech recognition (recordings, realtime transcription, live captions) and text translation cover 14 languages
(text translation also covers Tamil ta):
| Language | Code |
|---|---|
| English | en |
| Hindi | hi |
| Spanish | es |
| French | fr |
| Chinese (Mandarin) | zh |
| German | de |
| Arabic | ar |
| Brazilian Portuguese | pt |
| Russian | ru |
| Japanese | ja |
| Korean | ko |
| Italian | it |
| Dutch | nl |
| Vietnamese | vi |
Text-to-speech has ready-made voices in 6 languages (English en, Spanish es, French fr, German de, Brazilian
Portuguese pt and Hindi hi), and designed voices in all 14.
Language support by capability
Section titled “Language support by capability”| Capability | Lane / host | Input | Languages |
|---|---|---|---|
| Speech-to-text — batch (Hindi, other languages) | gpu-asrhq |
audio file | the 14 languages (auto-ID); task=translate → English text |
| Speech-to-text — fast file transcription | gpu-transcribe |
audio or video file | en, es, fr, de, pt, ru, it, nl; text or subtitles |
| Speech detection | gpu-transcribe (/v1/audio/segments) |
audio or video file | any (finds speech, no transcript) |
| Speech-to-text — realtime | gpu-asr (WebSocket) |
16 kHz PCM | the 14 as locales: en-US, hi-IN, es-ES, fr-FR, de-DE, pt-BR, ru-RU, ar-AR, zh-CN, ja-JP, ko-KR, it-IT, nl-NL, vi-VN |
| Live captions | gpu-live (WebSocket) |
16 kHz PCM | en, hi, es, fr (auto-detect or pinned); de, pt, ru, ar, zh, ja, ko, it, nl, vi pinned. Captions in the speaker’s language |
| Live translation | gpu-live or gpu-asr, then ai.ollalink.com (/v1/translate) |
speech, then text | speech in the 14, translated into any of the 14 or Tamil (Live captions) |
| Text translation | ai.ollalink.com (/v1/translate) |
text | the 14 languages and Tamil (ta), any direction (Translation) |
| Text-to-speech | gpu-tts / gpu-voice |
text | en, es, fr, de, pt, hi — 34 voices (Voices by language); the 14 languages with "voice": "design" (Designing voices) |
| Spoken numbers to digits | ai.ollalink.com (/v1/normalize) |
text | en, hi |
| Text embeddings | ai.ollalink.com (/v1/embeddings) |
text | — |
| Classification | ai.ollalink.com (/v1/classify) |
text | — |
Ask the API what is available
Section titled “Ask the API what is available”Every host answers GET /v1/languages (with your GPU API key; not counted as usage) with the languages it serves right now,
so your app can build its language and voice pickers from the API.
| Host | Lists | Flags per language |
|---|---|---|
gpu-asrhq.ollalink.com |
batch speech-to-text languages | accuracy_recognizer, numbers_to_digits |
gpu-transcribe.ollalink.com |
fast file transcription languages | numbers_to_digits |
gpu-live.ollalink.com |
live caption languages, with locale |
auto_detect, numbers_to_digits |
gpu-asr.ollalink.com |
realtime speech-to-text languages, with locale |
auto_detect, accuracy_reread, numbers_to_digits |
gpu-tts.ollalink.com, gpu-voice.ollalink.com |
speech languages with default_voice and voices |
— (voice details: GET /voices) |
ai.ollalink.com |
translation languages, any direction | normalize |
curl https://gpu-live.ollalink.com/v1/languages -H "X-NH-GPU-Key: $YOUR_GPU_KEY"# {"object": "list", "route": "wss /v1/translate/stream",# "data": [{"code": "en", "name": "English", "locale": "en-US", "auto_detect": true, "numbers_to_digits": true}, …]}Chat, coding and other models
Section titled “Chat, coding and other models”Chat and coding ("model": "chat" / "coding"), document OCR, classification and the default image model are always on;
vision, the second image model, audio generation, music and speaker diarization (up to 8 speakers) load when you call them.
See Chat and coding and Models.
Not offered
Section titled “Not offered”| Capability | Status |
|---|---|
| Speech-to-speech | sound-stream is paused (503); translate captions with /v1/translate and speak them with text-to-speech |
| Ready-made voices in Chinese, Arabic and the other languages outside the six | not offered — use "voice": "design" |
| Voice cloning from your own clip, the voice studio routes | paused — see Voice cloning |
| Kannada, Tamil, Telugu, Bengali speech on live captions | refused (close code 1008) |
| Urdu, Turkish | not offered; use hi for Hindi |
Choosing the right lane
Section titled “Choosing the right lane”- Captions of a live speaker? Use
gpu-live(Live captions); for your own UI with commit control, the realtime lane ongpu-asr. - Live translation of a meeting or class? Caption it with
gpu-liveand send each final toPOST /v1/translatewith"stream": true; optionally speak the translation with Text-to-speech in English, Spanish, French, German, Brazilian Portuguese or Hindi. - Transcribing English or European-language recordings, or need subtitles? Use fast file transcription
(
gpu-transcribe). - Hindi, Chinese, Arabic, Japanese, Korean or Vietnamese recordings, or English text from foreign speech? Use
gpu-asrhq.
Tested scope versus available languages
Section titled “Tested scope versus available languages”The lists above describe available API routes. They do not qualify every accent, number construction, grammar pattern, long meeting or shared-load workload. See Speech routing before planning a large event.
Need a language we don’t list? Tell your ollalink contact. Related pages: Speech-to-text, Text-to-speech, Voices by language, Live captions, Hindi with English numbers.