I know, the eternal thorn in our side. A few weeks ago I mentioned I was going to do a deep dive digging into transcription… and it got snowed under other things. I’m starting this new topic so we can do a round of “here’s how I do it” — please chime in with your thoughts on how you generate transcripts.
This is a classic example of one of the points of this community existing: Creating publicly readable, useful stuff from what we collectively use and know.
I use a variety of services…
I’ve sometimes used (although not for over a year now) Rev.com’s human transcription. It comes back as perfect as can be. (Unless it has french parkour terms in it… ¯\_(ツ)_/¯ )
I’m currently getting machine-generated transcripts out of Zencastr.com as part of my regular service. (I’m not on the free plan.) These are pretty good, and I whip through them spending a few minutes delete some “um.” and “uh.” transcript breaks, fixing names and other proper nouns. That’s how the transcripts for #listen come to exist.
I also have a Pro Otter.ai account which lets me upload audio files (in addition to all the ways everyone else uses Otter ;) I use this when I want a transcript of some podcast I just listened to when I want to write posts or whatever. If I record a podcast using Zoom (happens rarely, but it happens) then I’ll send the audio file over to Otter to get a transcript.
I’ve talked about switching to using Amazon’s transcription service. It’s insanely cheap. Like 2-cents/minute cheap. (For the first 250,000 minutes of usage a month. Beyond that, the price goes down.) But I’ve just not had the time to try it out, and Otter.ai and Rev.com when I need a hammer human, do a good-enough job.
I like Descript for transcripts. I’ve gotten used to the shortcuts, although there are things in there that are not very intuitive (overdubbing, for one), the transcripts are pretty straightforward.
After Craig mentioned the Amazon AWS transcription method at a recent meeting I could help check it out.
It seems pretty good. NOT intuitive to use, but not impossible either.
I’d didn’t really assess the quality of the transcript fully because I was kind of surprised I got it to work at all! I’ve never had AWS work for me before so I was expecting it to be impossibly difficult.
I’m using Descript at the moment so I don’t really need a new method, but next time Descript let’s me down - probably with my own Scottish accent - I know I have another option, and I have discovered that Descript allows you to import a transcript to replace theirs, so then I can still use it for edits and caption videos.
I’ve haven’t used it and can’t provide any firsthand experience, so can only mention that I have a friend that has used Trint for years and he’s very happy with it.
came up in the Akimbo POD11 course and I wanted to capture what I wrote in there, out here too…
It’s still early days for transcripts being 1st-class citizens in the podcasting technology.
It still feels to me like transcripts are best placed on our own platforms. (There’s no harm of course in posting it in our chosen podcast host’s interface/web-site.)
Today, I’m still thinking of transcripts as most-useful for people to find via search engines. (So Podtalk’s transcripts are posted inside Topics in the Podcaster Community, to attract search-engine traffic. Movers Mindset is, at this moment, not publishing transcripts anywhere.)
In the future, the podcasting technology is moving towards having the “transcript is at this URL” information embedded in the underlying RSS data feed for our shows. This is bleeding edge today. Eventually, we’ll put our transcripts somewhere—possibly on our podcast host’s web platforms, possibly on our own web sites—and we’ll add a URL to our episodes’ information. Then transcripts will be 1st-class citizens; visible in podcast players and everywhere that our audio is available.
Further into the weeds:
Similar tech is coming for “where’s the discussion” that goes with this episode? “where’s the video?” credits for this episode?
If you want to learn more about how that will work, check the following. (Nothing to do with me. Not to be confused with things you’ve possibly heard me say about podcast directories.) The “namespace” here is about namespaces in XML; the coding that our RSS feeds are currently written in.
Some of today’s options and possibilities will become tomorrow’s standard.
I think some kind of automation to the process will likely drive it forward. The visible automatic creation of captions is an example.
Apple Podcasts also are experimenting with creating transcripts so they can tag podcasts with topics determined by the content - but not really shouting about it - read an article on podnews.net and not making the tags or transcripts available.
The important question for now is what do your listeners need? Will creating and managing transcripts add to your overall purpose or just your workload?
This is a classic example of one of the points of this community existing: Creating publicly readable, useful stuff from what we collectively use and know. 