A Guth Labs publication ↗
Back to News

AWS adds a single-request API for Amazon Transcribe

A green glass waveform meets blank cream paper strips at a gold band.
AI-generated editorial illustration

Amazon Web Services introduced a synchronous API for Amazon Transcribe on November 20. The API, called TranscribeAudio, accepts a complete audio file and returns the transcript in the same response. That gives developers a third option alongside submitting a background job and opening a live audio stream.

AWS lists supported files from half a second to one hour. An application can send up to 25 MB directly or use an Amazon S3 file of up to 500 MB. The API can add word-level timestamps, separate speakers, identify the language and use a custom vocabulary. AWS says the price is $0.007 per minute of audio, billed by the second. Redaction of personally identifiable information costs extra and is not available in São Paulo.

The API is available in 13 AWS regions, according to the announcement. The change matters for applications that already have a complete recording: they can request a transcript without building a separate polling loop or maintaining a streaming connection.

Sources and citations

Revision history

Published .