Skip to main content
SpeechRelay

Turn overlapping speech into a transcript.

SpeechRelay turns streaming audio from several talkers into a transcript, with each voice kept in its own turn.

Try the free playgroundRead the docs

Requires sign-in. English speech only. The playground has limited availability, so a session may be refused.

Mara

Where shall we meet?

Jon

The café has a table outside.

Noor

By the river, if the weather holds.

API access is not open yet. Join the API waitlist to ask for access when it opens.

What it does

Available now

  • Streaming transcription

    Audio goes in over a streaming connection and transcript text comes back over the same stream.

    Available nowRequires sign-in. Limited availability.

  • Multi-talker transcription

    Speech from several speakers in one conversation is transcribed and told apart.

    Available nowRequires sign-in. The free playground handles up to eight speakers per session, four at a time.

Planned

  • Target-speaker transcription

    Given one voice to follow, SpeechRelay returns that speaker's words from a mix of voices.

    PlannedNot open yet. It is not available in the playground or the API.