01
Upload or Stream Audio
Send a recording for batch transcription, or stream live audio. Phone lines, VoIP, web, and mobile audio are all supported.
Speech-to-Text
Explore speech recognition for regional languages, dialects and multilingual conversations. Choose a language to see how Shunya Labs
supports your transcription workflow. For language availability and model selection, check the latest product documentation.
I'd like to check the delivery schedule.
मैं डिलीवरी शेड्यूल चेक करना चाहता हूँ।
配送のスケジュールを確認したいです。
Languages
31 published languages
Process
01
Send a recording for batch transcription, or stream live audio. Phone lines, VoIP, web, and mobile audio are all supported.
02
Lock a language code when you know it, or leave automatic language detection on when you do not. For clips shorter than about five seconds, locking the language is more reliable.
03
Horizon returns the transcript, with optional speaker labels, word timestamps, punctuation, and formatting in the same request.
04
Use the same speech-to-text API from your existing voice stack, for batch files or a live WebSocket stream.
Use Cases
Transcribe support calls in the customer's language, then extract intent and sentiment for agent assist and QA. See contact centre deployments.
Give a voice agent a transcript it can act on, including calls that switch languages mid-sentence. See voice agents.
Turn broadcasts, interviews, and recordings into captions and searchable text.
Capture multi-speaker meetings as a transcript with speaker labels, so notes and follow-ups start from what was actually said.
Add transcription to a product you already ship, from in-app voice input to archived call audio. Start from the speech-to-text product.
Questions
Horizon supports speech recognition across 204+ languages. This directory lists the regional language pages published so far. A missing card does not mean the language is unsupported. More language pages are added here when they go live.
Yes. Horizon streams transcripts for live audio, with a sub-500ms first token, and also transcribes uploaded recordings in batch. Phone, VoIP, web, and mobile audio are supported in both modes.
Yes. Leave language detection on when the language is not known in advance, or lock a language code when it is. On clips shorter than about five seconds, detection is less reliable, so setting the language explicitly is the better choice for short audio.
Yes. Horizon Code-Switch transcribes conversations that move between Indic languages and English without a separate language-detection pass. Japanese speech-to-text also handles Japanese-English switching in the same utterance. Each language page describes how that language is handled.
Get Started
Explore our speech recognition platform or contact our team.