API
Self hosted
On device (soon)
Speaker identification
Combined with our world-class speaker diarization technology and state-of-the-art voiceprints, speaker identification tracks recurring speakers across multiple podcast episodes or team meetings.
Overlapping speech detection
Overlapping speech happens all the time in spontaneous conversations (fast turn-taking, back-channeling, or interruption). Not only do we detect it, but we also attribute it to the right speakers.
Voice activity detection
Our world-class speaker diarization technology builds on top of state-of-the-art voice activity detection. Removes regions without speech, yet detects even the shortest words ("yes", "oh", "hmm").
Confidence score
Highlights the most complex parts of a conversation automatically. Use confidence scores to filter out the noisiest parts of your training data, or for human-in-the-loop use cases.
Speaker separation (soon)
Our world-class speaker diarization technology not only detects overlapping speech, it can also isolate overlapping speakers and return one separate audio stream per speaker.
