Saidwire demo · Voice agent
Talk to a voice agent.
Click start, allow the microphone, and talk. The agent hears you through the Saidwire speech-to-text API, then answers out loud. Use headphones or speakers, either works.
Up to 3 minutes per session.
How it works
1. Listen
Your microphone audio streams to our server. The Saidwire API turns it into words as you speak. When you pause, the agent asks for the final transcript and gets it back in tens of milliseconds.
2. Think
A hosted language model writes a short reply. The agent waits about a third of a second after you stop, so it does not cut you off mid-thought.
3. Speak
A hosted text-to-speech voice reads the reply back while it is still being written. Start talking over it and the agent stops to listen.
Built with Pipecat, the open-source voice-agent framework, using the Saidwire Pipecat plug-in. Times shown are measured on our server and do not include your own internet connection. Audio is processed in memory and never stored; transcripts are not saved. Demo sessions are limited per visitor.