VoiceBio

VoiceBio turns speech into a voiceprint — a compact representation of what a speaker’s voice sounds like, independent of what they said. Once you hold a voiceprint you can answer two questions:

• Is this the person they claim to be? StreamingVerify compares live audio against one voiceprint and returns a similarity score and a match decision.
• Which of these people is speaking? StreamingIdentify compares live audio against a list of voiceprints and returns the best match, or none.

Voiceprints come from StreamingEnroll, which reports how much more audio it needs before enrollment is complete. They can be stored and reused: to compare voiceprints you already hold, without sending audio again, use CompareVoiceprints, and to store them as numerical vectors for your own indexing, use VectorizeVoiceprints.

The engine runs on your own hardware, so the audio and the voiceprints stay inside your infrastructure.