Office 365 Exposed Episode #12

Discussing the State of Office 365

At 9am last Friday, the last day of the Ignite conference in Orlando, Paul Robichaux and I got together with Greg Taylor to discuss various topics loosely related to Office 365 in the latest episode of Office 365 Exposed, a podcast that we tape on an irregular basis when we are in the same place together for more than a day or so. You can subscribe to the podcast on iTunes and read Paul’s commentary on this episode on his blog.

Nine is a terribly early time to talk about anything, especially after being out late the night before at the conference’s attendee party at Universal Studios and then getting up early to fly in Paul’s Cessna over Orlando, so some of what we said might not make too much sense. At least, that was my conclusion after reading the transcript automatically generated by Stream when I uploaded the raw MP4 file given to us by the film crew a few minutes after we finished.

Video Transcripts

The idea behind automatic transcripts is that Stream is able to recognize words spoken on videos as it processes and publishes uploaded video content. Automatic speech recognition technology breaks the spoken word down into a series of captions that combine together to form the transcript. There’s a lot of heavy-duty technology here and only English and Spanish language videos are supported today.

But nice as it is to have an automatic transcript, the ability of the speech recognition engine to understand what is being said varies from very good to not so good. Take this interchange about the Teams background blur feature:

“you’d solve our resolve the full-grown blur yours as well but just enough“

“as an option to reverse that generative for way I’ve learned by interface“

“yeah, they’re nice I say that it that all actually that is a good point to make people aware of is”

Generating transcripts from videos is an advanced feature of Stream that’s part of Office 365 E5 (Stream Plan 2). And to be fair, it must be extraordinarily difficult to cope with three different accents (English, U.S., and Irish) from three grumpies, all of whom have no difficulty talking over each other.

In any case, you can edit the transcript to increase accuracy, and this is something that’s probably done by corporate marketing or internal communications people when important videos are posted to Stream for internal consumption. Fully correcting a transcript probably takes three times as long as a video, so you’d expect to spend at least two hours on this 43-minute extravaganza.

Meanwhile, the nice people at Practical365.com took the raw video and did a little editing to remove the bits before and after we taped the show and have posted the finished article for your viewing pleasure. The content didn’t improve, but perhaps you’ll be able to make more sense of what we said than Stream did.

---

We cover Stream in Chapter 7 of the Companion Volume for Office 365 for IT Pros. As Microsoft get through the migration of Office 365 Video to Stream, we might bring the topic back into the main book. Then again, we might not.