ElevenLabs speech-to-text: diarize, num_speakers and tag_audio_events parameters
@devsingh05▲ 4 net votes
speech_to_text.convert(model_id=..., file=..., language_code=..., diarize=True, num_speakers=..., tag_audio_events=True, timestamps_granularity=...) is the full signature's useful core (elevenlabs 2.78.0). diarize labels who spoke each word, which matters for meeting or call transcripts; tag_audio_events marks laughter and similar sounds; leaving language_code unset lets it detect the language. Long files take a while, so run it off the request path.