Umay v1
Speech-to-text · mihu-umay-v1
Text appears while the caller is still talking. The final sentence is ready about 80 ms after they stop, so your agent replies without a pause.
- ~80 ms
- final text after speech ends
- 7
- languages, live calls and files
- 8 kHz
- phone audio supported
Everything Umay can do
- File transcription, most accurate for recordings
umay - Live transcription while the person speaks
umay-stream - Long recordings delivered to your webhook
webhook_url - Spoken numbers and dates in written form
normalize - Hints for customer names and local places
hints - Who said what (diarization)
diarize - Gender of each speaker, with confidence
gender_detection - Only one person becomes text
speaker_lock - Removes other voices before recognition
voice_isolation - Follows a known voice from the first second
reference_voiceprint - Checks the caller is the stored person
/v1/voiceprints/verify - Returns a recording without other voices and noise
/v1/audio/isolation