Why You Should Never Drink & Transcribe
In another professional life, I was a government administrator. And of course, government is synonymous with meetings, and it always fell to some unlucky person to take notes.
One of my colleagues, who had spent years as an old-school secretary, was extremely adept at shorthand. It was this amazing, other-worldly language that enabled her to take notes almost faster than we were talking.
Fast-forward to today, and I don’t imagine there are too many people using shorthand in meetings. Although we have the tools for machines to take the notes for us (i.e. transcription), it is still not a widely used technology.
And it’s not just meetings that need transcription — it’s podcasts, calls, notes-to-self, webinars, customer service calls, interviews… and the list goes on.
There are two types of automation when it comes to transcription. There’s the automate-to-human approach, and fully automated AI transcription.
An example of an automate to human system is something like Rev.com.
Using Rev, you can just send over a sound or video file, and a hardworking transcriber will listen and type it out. The cool thing is that Rev connects with Zapier, so you can link it together with other services.
It could be as easy as dropping a file in Google drive, and with a few steps in a Zap have the transcription on your blog.
In my experience, it’s pretty accurate, and needs little to no adjustment. The downside is that it’s not instant, and the cost is relatively high at around USD $1.00 per minute.
On the other hand, a fully automated system like Temi or Rev.ai takes a recording and runs it through some pretty complex AI to deliver a fairly accurate transcription, with no humans involved.
Here’s an example of a short clip of audio (of yours truly), and below it, the transcription through Temi:
The thing is, this is an example in perfect circumstances — it’s quiet, there’s only me talking, and despite being Australian, I speak relatively clearly.
It becomes less accurate when it’s a recording of a group of people talking over each other, there are very strong accents, or you’ve had a few too many drinks. So, avoid drinking and transcribing.
That being considered, there is a risk that the transcription could come out a bit odd.
At USD $0.10 a minute, it’s a bargain (and at least worth testing out). And like Rev.com, it’s Zapier ready.
Rev.ai is functionally identical to Temi, but doesn’t offer the same front-end interface and usability. It’s designed more for developers building custom platforms that require a lot of transcription — think call centres or a high volume audio files.
Rev.ai costs less at USD $0.035 per minute, so if you’re a power user of transcription services, it might pay to put in the development time to make something more specialised. Of course, I’ve turned this into an Automation Hack.
So we’ve come a long way from shorthand, but I didn’t imagine we’d already be this far with AI transcription. It might still be a bit rough around the edges, but it is getting better scary fast.
tl;dr: Automated transcription is awesome. You can automate to a human transcriber, or let the machines do it.
