transcribe

To transcribe audio with speaker diarization, run the diarize model to produce diarized_json output. This generates timestamped text segments labeled by speaker, which you can save to organized job directories for meetings or interviews.

Yes, you can transcribe long audio files by using the chunking feature. The workflow validates file size and splits larger audio inputs into manageable chunks, processing them sequentially to ensure complete transcription without hitting size limits.

Speech-to-text conversion supports common audio formats for input and outputs text, json, or diarized_json. This allows you to export standard transcripts or structured JSON files with speaker segments for downstream analysis and publishing.

Yes, you need an OpenAI API key to convert recorded meetings into text. The transcription workflow validates your API key before processing to ensure access to the speech-to-text models required for generating accurate transcripts.

To label known speakers in an interview transcript, provide reference audio hints when running the diarize model. The transcription process uses these hints to accurately assign speaker labels within the diarized_json output for your specific speakers.

The best way to get a quick text transcript is using the default lightweight transcribe model. It processes short audio clips rapidly and outputs plain text, making it ideal when you need fast transcription without speaker labels.