Transcribe and Summarize a Call Recording
A 4-step recipe with RingCentral. We can help you get access.
Recipe · 4 steps
Transcribe and Summarize a Call Recording
with RingCentral
Transcribe a call recording with diarization and produce a conversational summary.
- Submit the recording for asynchronous speech-to-text. Diarization and punctuation are enabled so the transcript is returned as clean, per-speaker utterances rather than an undifferentiated word stream. RingCentral responds 202 with the job
- Poll the shared AI job status endpoint until the transcription job reports a terminal status. On success the response carries the transcript, the per-speaker utterances, and an overall confidence score.
- Feed the diarized utterances straight into conversational summarization. The utterance objects returned by speech-to-text already carry the speakerId, text, start, and end fields this operation requires.
- Poll the same job status endpoint for the summarization job and collect the finished summary.
What you'll need to have ready
- contentUri: Publicly facing URI of the audio to analyze. The RingCentral AI service fetches the media from this location.
- encoding: Encoding of the source audio. One of Mpeg, Mp4, Wav, Webm, Webp, Aac, Avi, Ogg.
- languageCode: Language spoken in the audio (e.g. "en-US").
- speakerCount: Number of speakers in the recording. Omit when unknown and let the service detect it.
- summaryType: Type of summary to compute. One of Extractive, AbstractiveShort, AbstractiveLong, AbstractiveAll, All.
For your developer
Arazzo workflow ringcentral:transcribe-and-summarize-recording, 4 operations: caiSpeechToText → caiJobStatusGet → caiSummarize → caiJobStatusGet. Specs on APIs.io.