Speech-native voice AI agents
Open-weight voice models with sub-second latency: conversational agents without proprietary locks.
The open speech-to-speech models (Fixie): sub-second conversational voice that you can self-host, the OSS answer to proprietary voice APIs.
Who it's for
Developers building real-time speech-to-speech voice agents without a pipeline of separate models.
Official links
Keep exploring
1 in catalog
Ultravox features
1 in catalog
Ultravox has a pricing
1 in catalog