Amazon Transcribe has been around since 2017, and the book-shifter-turned-cloud-vendor has since added support for 31 languages, including six that can be transcribed in real-time.
It isn't alone. Microsoft's Azure Cognitive Services will understand more than 40 languages (and variants).
Speech-to-text is a handy thing to have for generating transcripts or searching through audio content for key words or phrases. Amazon's latest wheeze is to spot the language being spoken, reducing the need for manual tagging or forcing a user down a certain path from the get-go.
The identification requires a minimum of 30 seconds of audio and works for all of Amazon's 31 supported languages, although the cloud giant admitted that accuracy was improved if the list was reduced.
Those coding apps using Amazon's Transcribe service will receive a score between 0 to 1 to indicate how confident the service is that it has the right language. Amazon also suggested the service could be used to simply spot the language and nothing else, with just 30-45 seconds of audio needed in order to keep costs down.
Available in America, Europe, the Middle East, and Asia-Pacific, Amazon does not charge extra for the service. Handy, since costs can swiftly mount up once you've burned through the Free tier of 60 minutes per month.
Those noting the arrival of Nuance DAX integration with Microsoft Teams will also be interested in Amazon Transcribe Medical, a HIPAA-friendly service aimed at transcribing the likes of patient-clinician conversations. The pricing, at $4.50 per hour, is a little more expensive than the $1.44 of Amazon's standard. ®