Hear the World.
Every Language.
The fastest speech-to-text API for India.
All 22 Indian languages, 200+ worldwide.
~100ms latency. Your data never leaves Indian soil.
Press the mic to speak
Built for Real Conversations.
From call centers to courtrooms. See how Vega STT transcribes real-world audio across industries.
Go Beyond Transcription.
Dual-engine STT with enterprise features — keywords boosting, PII redaction, utterance segmentation, async webhooks — all toggleable per request, all on Indian infrastructure.
Keywords Boosting
NewBoost domain-specific vocabulary — company names, products, jargon. KenLM hotwords for Vega, decoder bias for Dhara.
PII Redaction
DPDPAuto-redact Aadhaar, PAN, phone, card, email, UPI. Indian DPDP Act compliant.
Speaker Diarization
pyannoteIdentify and label different speakers automatically.
Utterances
NewSpeaker-turn-segmented output with paragraphs.
Interim Streaming
500msVAD-optimized WebSocket with partial results every 500ms. KenLM on final only.
Word Timestamps
+5msPrecise start/end times with confidence scores.
Async + Webhook
NewSubmit audio, get results via callback URL.
~100ms Speed
WarmBlazing fast warm latency on Indian infrastructure.
Sentiment
AsyncPositive, negative, or neutral in any language.
Dual Engine
Vega for live voice agents. Dhara for meetings, content, and word-level timestamps.
Data Sovereignty
IndiaAll processing on Indian infrastructure. RBI, SEBI, IRDAI compliant. Data never leaves India.
Every Indian Language. Native Scripts.
All 22 scheduled Indian languages with native script recognition. Trained on thousands of hours of real-world audio.
नमस्ते, मैं आपकी कैसे मदद कर सकता हूं
নমস্কার, আমি কিভাবে সাহায্য করতে পারি
வணக்கம், நான் எப்படி உதவ முடியும்
నమస్కారం, నేను ఎలా సహాయం చేయగలను
नमस्कार, मी कशी मदत करू शकतो
નમસ્તે, હું કેવી રીતે મદદ કરી શકું
ನಮಸ್ಕಾರ, ನಾನು ಹೇಗೆ ಸಹಾಯ ಮಾಡಬಹುದು
നമസ്കാരം, എനിക്ക് എങ്ങനെ സഹായിക്കാം
ਸਤ ਸ੍ਰੀ ਅਕਾਲ, ਮੈਂ ਕਿਵੇਂ ਮਦਦ ਕਰ ਸਕਦਾ ਹਾਂ
ନମସ୍କାର, ମୁଁ କିପରି ସାହାଯ୍ୟ କରିପାରିବି
سلام، میں کیسے مدد کر سکتا ہوں
नमस्ते, मैं आपकी कैसे मदद कर सकता हूं
নমস্কার, আমি কিভাবে সাহায্য করতে পারি
வணக்கம், நான் எப்படி உதவ முடியும்
నమస్కారం, నేను ఎలా సహాయం చేయగలను
नमस्कार, मी कशी मदत करू शकतो
નમસ્તે, હું કેવી રીતે મદદ કરી શકું
ನಮಸ್ಕಾರ, ನಾನು ಹೇಗೆ ಸಹಾಯ ಮಾಡಬಹುದು
നമസ്കാരം, എനിക്ക് എങ്ങനെ സഹായിക്കാം
ਸਤ ਸ੍ਰੀ ਅਕਾਲ, ਮੈਂ ਕਿਵੇਂ ਮਦਦ ਕਰ ਸਕਦਾ ਹਾਂ
ନମସ୍କାର, ମୁଁ କିପରି ସାହାଯ୍ୟ କରିପାରିବି
سلام، میں کیسے مدد کر سکتا ہوں
নমস্কাৰ, মই কেনেকৈ সহায় কৰিব পাৰোঁ
प्रणाम, हम कोना मदद करी
नमस्कारः, अहं कथं साहाय्यं करोमि
ᱡᱚᱦᱟᱨ, ᱟᱹᱧ ᱪᱮᱫ ᱞᱮᱠᱟ ᱜᱚᱲᱚ ᱫᱟᱲᱮᱭᱟ
آداب، بٕہ کِتھ مدد کٔرِ ہیکو
नमस्ते, म कसरी मद्दत गर्न सक्छु
سلام، مان ڪيئن مدد ڪري سگھان ٿو
नमस्कार, हांव कशें मदत करूं
नमस्ते, मैं किदाँ मदद करी सकदा
ꯈꯨꯔꯨꯝꯖꯔꯤ, ꯑꯩ ꯀꯔꯝꯅ ꯃꯇꯦꯡ ꯄꯥꯡꯕ
मोजां, आं माबोरै हेफाजाब होनो
নমস্কাৰ, মই কেনেকৈ সহায় কৰিব পাৰোঁ
प्रणाम, हम कोना मदद करी
नमस्कारः, अहं कथं साहाय्यं करोमि
ᱡᱚᱦᱟᱨ, ᱟᱹᱧ ᱪᱮᱫ ᱞᱮᱠᱟ ᱜᱚᱲᱚ ᱫᱟᱲᱮᱭᱟ
آداب، بٕہ کِتھ مدد کٔرِ ہیکو
नमस्ते, म कसरी मद्दत गर्न सक्छु
سلام، مان ڪيئن مدد ڪري سگھان ٿو
नमस्कार, हांव कशें मदत करूं
नमस्ते, मैं किदाँ मदद करी सकदा
ꯈꯨꯔꯨꯝꯖꯔꯤ, ꯑꯩ ꯀꯔꯝꯅ ꯃꯇꯦꯡ ꯄꯥꯡꯕ
मोजां, आं माबोरै हेफाजाब होनो
Made in Bharat. For World.
Your data never leaves India. All processing happens on Indian infrastructure with full data sovereignty compliance.
Data Sovereignty
All audio data processed and stored within India. Compliant with Indian data localization requirements. Zero cross-border data transfer.
INR Pricing
Pay in Indian Rupees with no forex markup. Transparent pricing at ₹0.25/min. Volume discounts for enterprise customers.
Indian Infrastructure
Servers in Mumbai and Delhi for sub-50ms latency. Optimized for Indian network conditions. 99.9% uptime SLA.
Developer-first. Production-ready.
Vega is built for rapid prototyping and seamless integration. Ship voice-enabled products with a simple API, robust SDKs, and enterprise-grade infrastructure.
curl -X POST https://api.thinnest.ai/stt/transcribe \
-H "Authorization: Bearer thns_sk_..." \
-F "audio=@recording.wav" \
-F "language=hi"REST API
Integrate Vega directly into your product with simple, well-documented endpoints.
SDKs
Python, Node.js, and cURL — get started in under 5 minutes.
Playground
Experiment with real transcription instantly in your browser. Test any language.
Enterprise Grade
Data residency in India, encrypted at rest.
Audio never leaves Indian infrastructure. Full sovereignty.
pip install
vega-sttHow Vega STT Compares
| Feature | Vega STT | Deepgram Nova-3 | Sarvam Saarika | AssemblyAI | OpenAI Whisper |
|---|---|---|---|---|---|
| Languages | 200+ | 36 | 10 | 18 | 99 |
| Indian Languages | 22 | 1 | 10 | 1 | 7 |
| Latency | ~100ms | ~300ms | ~500ms | ~400ms | ~2s |
| Price/min | ₹0.25 | ₹0.58 | ₹0.50 | ₹0.55 | ₹0.50 |
| Self-hosted | |||||
| Data Residency | |||||
| Speaker Diarization | |||||
| Word Timestamps | |||||
| Auto-formatting | |||||
| Sentiment Analysis | |||||
| Keywords Boosting | |||||
| PII Redaction | |||||
| Utterance Segmentation | |||||
| Async/Webhook | |||||
| VAD Streaming |