Power best in class voice agents

Ultra-fast and ultra-accurate streaming STT built for voice agents. Get final transcripts in a 307 ms median (Pipecat open benchmark) and entity-aware endpointing.

Start building Read the docs

Listen and speak

Two APIs for your voice agent

Universal-3.6 Pro Realtime hears every caller. Text-to-Speech gives your agent a voice callers stay on the line for.

Listen · Live now

Universal-3.6 Pro Realtime

The STT layer for your cascading voice-agent architecture. Works natively with your preferred orchestrator, with Sync Speech-to-Text available for finished utterances.

View integration docs

$0.45 /hr

Transcription only, unlimited concurrent streams

Teams already using LiveKit, Pipecat, or Vapi as their orchestration layer
Teams running cascading architectures: STT -> LLM -> TTS
High-scale deployments where margin and full control matter
Complex workflows with RAG, custom tooling, or proprietary LLMs
HIPAA, SOC 2 - bring your own compliance infrastructure
Sync Speech-to-Text API for teams running their own VAD and turn detection

Speak · Live now

Text-to-Speech API

Trained on real conversational speech, so your voice agents sound human, get names and numbers right, and keep callers on the line.

Explore Text-to-Speech

$15 /1M chars

Billed on input characters, cloned voices included

About 90 ms time to first audio, so your agent answers inside the turn
Order numbers, prices, dates, and confirmation codes come out exactly as written
4,000+ voices, plus instant voice cloning from a short clip
English, Spanish, French, German, Italian, and Portuguese
No spend commitment to unlock the concurrency you need
Streaming and HTTP APIs, with word-level timestamps

Start building

Explore our comprehensive docs with integration guides and best practices to optimize accuracy and latency for your application.

Read the docs

What each API handles

Universal-3.6 Pro Realtime

Streaming speech-to-text

Text-to-Speech API

Conversational text-to-speech

Role in your agent

Listens: turns caller audio into text

Speaks: turns your LLM reply into audio

Built for voice agents

Speed

307 ms median time to final transcript (Pipecat open benchmark)

About 90 ms time to first audio

Accuracy

Entity-aware endpointing

Order numbers, prices, dates, and codes read exactly as written

Languages

32 languages

English, Spanish, French, German, Italian, and Portuguese

LLM

Bring your own, or route through LLM Gateway

Integrations

LiveKit, Pipecat, custom WebSocket, Twilio SIP

LiveKit, Pipecat, streaming WebSocket, HTTP

Scaling

Unlimited concurrent streams

No spend commitment to unlock concurrency

Pricing

$0.45/hr — STT only

$15 per 1M characters

Integrations

Ready to plug into your voice-agent stack

Pre-built integrations with step-by-step docs enabling quick implementation without disrupting existing workflows.

Join 200K+ developers building new experiences with voice data

If you have an hour of content, the difference between 99% accuracy and 97% accuracy, it's a lot of time for that person to review. So you could cut down their workflow from taking half an hour, to 20 minutes, to 15 minutes — it's huge, right?

Joshua Grossberg

CTO, Kapwing

Investments in STT improvements always pay for themselves, since it is such a critical building block of the voice pipeline.

Lindsay Liu

Co-Founder & CEO, Super

AssemblyAI's managed API endpoint and diarization won me over — something Whisper couldn't provide.

Josh Mohrer

Founder, Wave.co

We needed a provider that could scale with us — offering unlimited concurrent streams, fair pricing, and responsive support.

Mark Barbir

CEO, Earmark

The transcription accuracy, reliability, and speed of AssemblyAI's API have greatly enhanced our operations.

Raj Shankar

SVP Product, Calabrio

Our free to paid conversion rate doubled after implementing AssemblyAI.

Colin Treseler

Founder & CEO, Supernormal

The accuracy was strong, but the great documentation and unique models like Auto Chapters and Sentiment Analysis is what really won us over.

Nathan Webb

Product Manager, Aloware

Calls are and will remain a pertinent part of the customer service journey. Customer service isn't moving entirely to chatbots or chat interactions.

Dr. Shane Lynn

CEO, EdgeTier

Common questions

: Run a cascading pipeline (STT → LLM → TTS) in an orchestrator like LiveKit or Pipecat, or your own WebSocket client. Universal-3.6 Pro Realtime is the speech-to-text layer that hears the caller, and the Text-to-Speech API gives your agent its voice. If your orchestrator already handles VAD and turn detection, the Sync Speech-to-Text API transcribes each finished utterance in a single HTTP request with no streaming connection to manage.
: Universal-3.6 Pro Realtime works with LiveKit, Pipecat, Python SDKs, Twilio SIP, and custom WebSocket clients, so it drops into the voice-agent stack you already run.
: Streaming and HTTP APIs with about 90 ms time to first audio, 4,000+ voices, instant voice cloning, word-level timestamps, and six languages: English, Spanish, French, German, Italian, and Portuguese. Processing runs in the US or the EU at the same price, and there is no spend commitment to unlock concurrency.
: Universal-3.6 Pro Realtime is $0.45/hr for speech-to-text, with unlimited concurrent streams and no concurrency fees. The Text-to-Speech API is $15 per 1M characters. Run your LLM through LLM Gateway or bring your own.
: If you already use an orchestrator like LiveKit or Pipecat with your own LLM and TTS providers, swap in Universal-3.6 Pro Realtime as the STT layer in minutes. Then try the Text-to-Speech API on a slice of your traffic and compare it with your current voice.

You may also be interested in

Voice agents for real-time agent assist Voice agents for meeting intelligence & AI notetakers Voice agents for field service operations Voice agents for customer support & contact centers Voice agents for sales & call intelligence Voice agents for healthcare Voice agents for financial services Voice agents for AI medical scribe & ambient documentation Voice agents for multilingual global support

Unlock the power of voice intelligence

Get started Learn more