Speech-to-text API for next-level apps
Convert speech to text with unmatched accuracy, ultra-low latency, and enterprise scalability. Deepgram’s speech-to-text API powers everything from transcription and analytics to real-time, human-like voice agents.
See Deepgram in action
Loading form


Leading voice teams build with Deepgram
Flux STT. One API. One conversational speech model. Ten languages.
Build voice agents that understand, adapt, and respond in real time across English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch.
Conversational speech recognition, now multilingual.
- Ultra-low latency conversational speech recognition, now global Built for real-time interaction, with turn detection and interruption handling across 10 languages
- Monolingual-grade accuracy with real-time language control Language hints, auto-detection, and native mid-utterance code-switching
- Build once, deploy globally One model, 10 languages, one API. No routing or per-language infrastructure

Transcribe in real time. Deploy anywhere. Scale effortlessly.
Lightning-fast transcription that doesn't compromise. Convert your most complex audio to text with best-in-class accuracy in seconds, not minutes.
API-first.
Integrate Nova-3 or Flux STT into your stack in minutes. Transcribe in real-time, or an hour of pre-recorded audio in <12 seconds.
Enterprise-grade security.
Compliant with SOC2, HIPAA, GDPR, and more. Regardless of what's in your speech data - or where it's hosted - Deepgram keeps it secure.
Cloud, on-prem, or VPC.
Your infrastructure, your rules. Flexible deployment options tailored to your infrastructure needs - choose the deployment method for you.
Nova-3 isn’t just an upgrade - it’s a leap forward.
- 54% accuracy lead over competitors in streaming data.
- 47% accuracy lead in pre-recorded transcription.
- 2-5X cheaper than legacy solutions.
- Near-zero latency for real-time AI.

50+ languages and dialects to choose from
Built for a global audience. With more than 50 languages and dialects supported, Deepgram adapts to your users - delivering accurate transcriptions no matter the language or locale.

Customization At your Fingertips
For years, training a speech AI model meant long development cycles and huge upfront costs. Nova-3 changes everything.
- Self-serve customization - Tune the AI to your business without hiring ML engineers & achieve near-perfect understanding of unique terminology out of the box.
- New keyterm prompting - Teach Nova-3 industry-specific terms instantly. Boost accuracy for up to 100 key terms.
No more waiting for custom models. No more compromises. Just next-level transcription, instantly.

AI Speech-to-Text That Thrives in the Real World
Most ASR models break down when faced with messy, real-world conditions. Nova-3 was built for them.
- Far-field speech – Even at a distance, Nova-3 delivers. From air traffic control to drive-thrus, get crystal-clear transcripts.
- Real-time multilingual conversations – Whether it's 911 dispatch or customer support, Nova-3 detects and transcribes multiple languages instantly.
- Hyper-specific industry keywords – Medical transcription, banking, eCommerce? Nova-3 handles it all without custom training.
- Number sequences & numeric entities – From patient IDs to credit card transactions, Nova-3 delivers unmatched accuracy.

Powerful Speech-to-Text. All the Features. No limits.
Designed for precision, security, and adaptability, our advanced features optimize transcription accuracy, context awareness, and seamless enterprise integration.
- Docs
Keyterm prompting
Instantly improve Keyword Recall Rate (KRR) for important keyterms or phrases up to 90%
- Docs
Filler words
Filler Words can help transcribe interruptions in your audio, like "uh" and "um".
- Docs
Smart formatting
Smart Format improves readability with punctuation and paragraphs.
- Docs
Diarization
Diarize detects speaker changes and labels each word in the transcript.
- Docs
Numerals
Numerals converts written numbers to digits (e.g., "one hundred" to "100").
- Docs
Redaction
Deepgram's Redaction removes sensitive information from your transcripts.

Ready to Build?
Use Deepgram with your existing tech or on its own—whatever fits your needs. Fill out the form to discover how our leading speech-to-text technology can help scale your products.





See Deepgram in action
Loading form














