# Ultravox

> Real-time voice AI platform with speech-native models for building and scaling conversational voice agents.

Ultravox is a real-time voice AI platform that trains speech-native models and runs them on dedicated, purpose-built infrastructure. Unlike traditional voice AI systems that convert speech to text before processing, Ultravox processes audio directly, preserving paralinguistic signals like tone, cadence, and pitch while reducing latency. The platform enables developers to build voice agents that can speak, listen, and interact in real time.

- **Speech-Native Model** - Ultravox v0.7 processes audio directly without transcription, achieving 91.8% on Big Bench Audio without reasoning and 97% with thinking enabled, making it state-of-the-art for voice intelligence.

- **Low Latency Architecture** - By managing the full inference stack and infrastructure in-house, Ultravox eliminates waiting on external LLMs or shared inference pools, delivering fast response times for natural conversations.

- **Dynamic Endpointing with UltraVAD** - Neural VAD model predicts conversation states and turn-taking by recognizing when users finish speaking, typical pause patterns, and distinguishing thoughtful pauses from turn endings.

- **Developer-Friendly APIs** - REST APIs enable easy integration, with powerful SDKs available for every major platform across web and mobile applications.

- **Built-in Telephony Support** - Native integrations with major telephony providers allow seamless deployment of voice agents for phone-based applications.

- **Custom Voice Clones** - Create personalized voice experiences with custom voice cloning capabilities, with limits varying by plan tier.

- **RAG Support with Corpora** - Build knowledge-enhanced voice agents using retrieval-augmented generation with configurable corpora.

- **Outbound Call Scheduler** - Pro and Enterprise plans include scheduling capabilities for automated outbound calling campaigns.

To get started, sign up for a free account at app.ultravox.ai. The platform includes 30 minutes of free calls and unlimited playground access for experimentation. Developers can integrate using the REST APIs and SDKs, configure voice agents with custom tools, and scale from experimentation to production with flexible pricing at $0.05 per minute.

## Features
- Speech-native AI model processing
- Real-time voice conversations
- Dynamic endpointing with UltraVAD
- Custom voice cloning
- RAG support with corpora
- Outbound call scheduler
- Telephony integrations
- REST APIs
- Web and mobile SDKs
- Unlimited playground calls
- No surge pricing

## Integrations
Telephony providers, SIP

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Freemium — Free tier available with paid upgrades

## Version
v0.7

## Links
- Website: https://www.ultravox.ai/
- Repository: https://huggingface.co/fixie-ai
- EveryDev.ai: https://www.everydev.ai/tools/ultravox