Transcription delivers accurate and customizable speech-to-text transcription. You can convert audio and video conversations into text in real-time or after the conversation has ended. speech-to-text features include:

  • Domain specific speech-to-text transcription models for mobile calls and video calls for state-of-the-art accuracy.

  • Accurate paragraphs and sentences with punctuation such as commas, periods, and question marks.

  • Support for multiple file formats like markdown .md and SubRip Text .srt.

Customizable Speech-to-Text features:

  • A custom vocabulary helps speech-to-text recognize specific words or phrases that are more frequently used within a context. If your conversations often include the word sell. When Speech-to-Text encounters the word sell, you want it to transcribe the word as sell more often than cell. In this case, you can use speech adaptation to bias Speech-to-Text to recognize sell. Enable the customVocabulary request parameter as described in the Streaming API.

  • Speaker Diarization connects individual speakers with their statements. Using standard speech-to-text transcription results in very accurate diarization, but not 100%. To achieve nearly 100% accuracy, generate speech-to-text transcripts using audio streams and pass the audio files with one speaker per channel. For more information, see Apply Speaker Separation to Async Files.

  • You can identify action phrases within a transcript to generate a list of action items from a conversation. For details, see the Conversations API Retrieve Action Items by conversation ID request or React Elements.


This request requires an access token, as described in Authenticate.

Get speech-to-text transcripts

This section describes how to get speech-to-text transcripts from a conversation. This request requires a conversation ID. You receive a conversation ID when you process a conversation.

To get a transcript of your conversation, use the following operation:


To make this request from the API reference, see Get messages.