In October, Microsoft unveiled a transcription function — Transcribe in Word — designed to let customers leverage the ability of the cloud to transcribe audio. After almost a yr in growth, Transcribe in Word is now usually obtainable in U.S. English for free of charge to present Microsoft 365 subscribers. It will come to Android and iOS later this yr.
You might say Microsoft is late to the occasion — speech-to-text is hardly novel, in spite of everything. But Microsoft undertaking supervisor Dan Parish says the corporate is “uniquely positioned” to supply a one-stop store for transcription. “You don’t have to worry about fussing around with different Windows apps,” he stated throughout a briefing with reporters. “What we’re trying to do with all of our investments in the natural user interface space — whether they have touch or voice, you name it — is enable everyone to work in the way that’s best for them so that they can be more effective, they can spend less time and energy creating the best work, and they can really focus on what matters most.”
Microsoft 365 subscribers utilizing Edge or Chrome will now see a Dictate menu beneath the Home tab after they create a brand new Word doc from Office.com. Selecting Transcribe will begin a recording, which will be paused at any time, whereas hitting the “Save and transcribe now” button will ship the recording to the Azure cloud for transcription. Prerecorded recordsdata in .wav, .mp4, .m4a, and .mp3 codecs will be uploaded through the brand new Upload audio choice.
Transcripts from recordings and uploaded audio seem within the Transcribe pane as soon as the transcription course of is full — shortcuts let customers rapidly insert sections or all the transcript into the Word doc. The time it takes to generate a transcript relies on web velocity and audio file measurement, which is restricted to 200MB and 5 hours per thirty days for recorded audio; uploaded audio is limitless. (Microsoft says it’s contemplating including choices to increase the previous cap sooner or later.) Recordings are saved within the Transcribed Files folder on OneDrive, the place they are often renamed or deleted, and there’s an enhancing software that can be utilized to alter a speaker identify for a bit, change all occurrences of that speaker label to a reputation, or repair names and typos.
When requested concerning the privateness implications of Transcribe in Word, Parish stated Microsoft doesn’t retain recordings or transcription outcomes however as a substitute shops them in customers’ private OneDrive folders. Recordings are solely despatched to the Azure backend to carry out transcription, not for any type of evaluation. Parish additionally claimed the speech recognition fashions underpinning Transcribe in Word have been educated on a “diverse” information set to make sure they acknowledge a variety of female and male audio system, together with these from totally different ethnic backgrounds.
Beyond transcriptions, Word on the net now acknowledges primary voice instructions like “start list,” “start numbered list,” “add italics/bold/underline,” and “add ellipses/ampersand/percent sign.” The full record lives within the assist panel, the place it may be checked with out having to click on away from the transcription mode.
The instructions are courtesy of Dictate, an add-on Microsoft retired final October in favor of a local Office 365 net and cell integration. Dictate helps 29 spoken languages, real-time translation to 60 languages, and two modes of punctuation. It additionally acknowledges pure language instructions like “add dot dot dot” (for ellipses), “pause dictation,” and “add comment,” in addition to casual instructions like “insert smiley face/heart emoji.” (Microsoft says Hindi, Korean, Russian, Polish, Thai, extra Spanish, and extra Chinese locales are on the best way.)