AI Aulis

AI NEWS IN ENGLISH · 21 November 2026

The most important AI news every morning. AI writes and checks the stories against the original sources – check important facts in the sources.

← Issue 21 November 2026Paper

Amazon Transcribe now returns an audio file’s transcript in a single call

AI-generated contentWritten by AI

aws.amazon.com, docs.aws.amazon.com · 7 sources · Sources published: 20 November 2026 · Some source publication dates are unknown

AWS’s new TranscribeAudio API accepts a complete audio file of up to one hour and returns the transcript in the same response, with no batch job to monitor. It costs $0.0070 per minute of audio.

Amazon Web Services (AWS) announced a synchronous TranscribeAudio API for its Transcribe speech recognition service on 20 November. Synchronous means the application sends a request and gets the result back in the response to the same call.

Until now, Transcribe has offered two ways to transcribe. In batch transcription, the application starts a transcription job and polls its status until the job finishes. The code examples in AWS’s documentation check the status in a loop every 5–10 seconds. In streaming, the application keeps an HTTP/2 or WebSocket connection open. Text arrives as partial results, which can still change as the speech continues. AWS said the new API needs no job to start, no status monitoring and no connection management.

The API does not replace streaming. It takes a complete audio file, so no text appears while someone is speaking. AWS suggests it for voice assistants, agent-assist tools and features in which a user asks for a single clip to be transcribed. The company said the transcripts are as accurate as those from batch transcription, but it gives no comparison figures.

The API supports word-level timestamps, language identification, speaker diarization and channel identification for up to two channels. It also formats numbers, dates, amounts of money and phone numbers. Users can pass a custom vocabulary and a vocabulary filter directly in the request, with no separate preparation step.

Audio can be sent inline with the request (up to 25 MB) or from the Amazon S3 storage service (up to 500 MB). Recordings can last between 0.5 seconds and one hour. The file formats are the same as for batch transcription: WAV, MP3, FLAC, M4A, MP4, Ogg, WebM and AMR. The transcript comes back in the response or is written to S3.

The price is $0.0070 per minute of audio, and AWS bills per second. The price includes speaker diarization, language identification, custom vocabulary and channel identification. Two-channel audio is billed once, based on the file’s duration. Personal data redaction costs extra.

Supported languages are English, Spanish, French, Italian, German, Japanese, Portuguese, Korean and Chinese. Finnish is missing, although it appears among the streaming API’s language codes. In Europe, the API is available in the Ireland, London and Frankfurt regions. It is also available in the US in the Northern Virginia, Ohio and Oregon regions, in Canada (Central), in the Mumbai, Singapore, Sydney, Tokyo and Seoul regions, and in São Paulo. Personal data redaction is unavailable only in the São Paulo region.

Report an error in this story