Addicted to novelty since 2001

Transcribing Podcasts and Other Audio for Fun, Profit and SEO

For one of our clients, we’re doing an informal series of audio interviews with tech gurus and web geeks about how they manage and back up their digital life. There are various marketing angles on this little project, but one is search engine optimization.

The first interview was with man-about-the-web Chris Pirillo. We talked for about 15 minutes over a dodgy Skype line. I had it transcribed by one of our contractors, and it turned into 2500 words of text.

That’s 2500 words of relevant text that’s reasonably rich with keywords. Assuming a going low-end rate of $25 per blog post, and 250 words a post, that’s $250 of text for the price of transcription and, all-in, about an hour of my time.

Exploring My Transcription Options

Yesterday I did the second interview in the series with the incomparable Vanessa Fox. I figured I could get the transcription done for less money, and I didn’t want to torture our contractor. I started shopping around for options. I also asked on Twitter, and these were the options that I came up with. The interview is 18 minutes, so I’ll include the pricing for that duration where available (all amounts, presumably, are in US dollars):

Then Avi reminded me of Andy Baio’s experiments in transcription with Amazon’s Mechanical Turk project. Andy has written up thorough instructions on how to outsource your transcription in chunks to workers in the system.

I decided to give Turk a try. It’s a little more work than just sending off the file to one of the aforementioned services, but I’ve always been charmed by Turk’s micro-task model. I divided the transcription into three six-minute chunks and posted the job. I offered $6 per chunk, or $1 per minute. I’m sure I could have gotten it done for less, but I’m looking for inexpensive, not cheap as humanly possible.

Mechanical Turk-Powered Transcription

It was mostly a success. The turn-around time was ridiculously quick. Despite my specifying a turnaround of a week being acceptable, I got all three transcriptions back in less than six hours. The quality seems excellent, though I don’t have a ton of experience with transcription to compare.

Despite my best efforts to ensure it didn’t happen, two workers transcribed the same section. Thus, I had to re-post a section to get complete results. And, of course, I have to paste the three chunks of text together and do some proofing. After Amazon’s fees, I paid $19.80 for 18 minutes of transcription.

Next time, I may just give CastingWords a try, as it would save me time and the difference in price is pretty marginal. There’s some interesting discussion on Andy’s blog about the ethics of pricing jobs on Turk. I was interested to learn that 75% of Turk workers are American, and that they have a variety of motivations for working on the service.

Posting Transcripts for Their SEO Benefits

I wondered on Twitter why more people don’t have their podcasts transcribed. The general response was that, for the hobbyist, it was too expensive. I guess it depends on if and how podcasters are monetizing their sites.

If you can get reasonably good transcription for $0.75 per minute, then you’ve got about 3700 words (my two transcriptions average out to 185 words a minute) to post on one or more pages for $15. Can you generate more than $15 from advertising on those pages over, say, two years? Maybe, maybe not, but it’s worth considering if you’re relying on advertising (and particular Google AdWords and the like) for revenue.

In any case, if they care one iota about SEO, a company has no excuses for not posting transcriptions of audio and video content. It’s not world-changing, but it’s another few steps in the marathon that is marketing.

UPDATE: Speaking of Mechanical Turk, Andy’s fruitful link blog points to the Turker’s Gospel, where Turk workers (‘Turkers’) rewrite Matthew, Mark, Luke and John in their own words.

16 Responses to “Transcribing Podcasts and Other Audio for Fun, Profit and SEO”

  1. Avi Bryant

    I assume you linked my name to the Dabble DB blog because that’s what comes up first when you google me, which is true but a little annoying. In an effort at SEO, can I request that you change the link to http://avibryant.com instead? :)

    Avi

  2. Duane Storey

    It’s a good experiment, but I’m not entirely convinced having a full transcript online is amazingly more SEO relevant that having a properly targeted preamble to a podcast or what-not. My mind’s still open here, but I’m still on the fence.

  3. darren

    Avi: Yep, that and laziness. I’ve changed that link.

  4. darren

    Duane: Yeah, I’m no SEO expert, so I wouldn’t know. It just seems to me that three pages with 1000 words on them plus your preamble is going to beat one page with just the preamble.

    I had to struggle to generate this example, but the transcription, in theory, attracts some relevant long tail searches that you otherwise wouldn’t get. This is a lousy example, but you get the idea:

    http://bit.ly/40UrQ7

  5. alanah

    Thanks for the excellent post with some great ideas. I had been following the Twitter dialogue on the topic as well.

    I happen to do transcription on occasion for a few outlets, and one issue that comes up (that you don’t mention) is ‘cleaning it up’. That is, journalists frequently clean up quotes from sources who may not be great speakers (i.e. they “hmmm” and “ahh” a lot) or for whom English is not their first language. Because of requirements of the clients I’ve dealt with, this issue comes up frequently. I’m usually transcribing interviews.

    The result is that it probably takes me about 20% longer than a straight-forward transcription and thus costs a bit more. On the other hand, because I’m only transcribing on a topic area I know very well, I might also be more accurate at times because I can understand the idioms and subject-specific language being used.

    I suppose it all depends on the priorities of the transcription and client in question.

  6. Duane Storey

    Yah, hard to say. There’s also the possibility that having more content would defocus the relevant keywords from a search engine crawler perspective. We should come up with an experiment to test it out using Google analytics and organic traffic. I.e. create two pages with similar URL structures, linked from the same page, one with a preamble only and one with the full transcription. Simmer for a few months and enjoy.

  7. DaveO

    No doubt podcast transcription is a worthy plan when seeking search engine love.

    Since i have no interest in doing it myself (especially after completing all the other steps required to make a killer show), nor do i want to spend bread to have my missives transcribed, the best option for me would be to ask listeners to do it.

    Since i don’t really care, i just put most useful terms and phrases in the show description which also goes into the blog post. A short paragraph of goodness rather than a tome of ramblings wins for me.

    In the future, perhaps my throngs of enthused fans can transcribe my hundreds of shows to share with future generations of student of counter-culture hi-jinx, spoken word literature, Canucks hockey and web community stuff.

  8. Ryan Cousineau

    I’d like to point out that Raul the bird said on Twitter yesterday that he has used Dragon NaturallySpeaking to do transcription. Apparently the $200 Preferred version will transcribe from a recording, and while I haven’t used it, it beats the typical $1/minute contract-pricing after you have about 4 hours of transcripts.

  9. darren

    Ryan: True, I left out Dragon. In part, I did so out of a long held (though earned through plenty of experimentation) bias about speech recognition: we’re not there yet.

    However, more practically, I don’t think Dragon will work for interviews, or for anything but dictation by a single speaker. As Raul pointed out, Dragon needs to be ‘trained’ for a particular voice, so I can’t imagine that it would work for a series of different speakers.

  10. Raul

    The way you’d work it with interviews is if you (the interviewee) dictate the response of the other person so that Dragon is already calibrated to your voice. More complicated, true, and it wouldn’t work for different speakers.

  11. Alexander Kandakoff

    Thanks for the post the services are indeed iinteresting..here in Ukraine we have nothing like this :) I wonder if you know any sites where you can send a text and they will record an audio for you.

  12. Jason

    I’m a bit late to the conversation, but this one particularly intrigued me because I work as a transcriptionist on a part-time basis. For 18 minutes, I would charge about $23 CAD, no minimum.

    I saw some really interesting stuff done by ZeFrank.com with regards to transcriptions a couple years ago. He had a lot of loyal and eager viewers and essentially crowd-sourced his transcripts to them. He had over 300 transcripts completed for ‘the show’. It’s a pretty incredible feat…

    http://www.zefrank.com/wikidata/index.php?title=Category:the_show_transcripts&from=the+show%3A+10-17-06

  13. Nathan Ketsdever

    Fantastic article. Waxy.org had a post (I believe it was 2 years ago, but I’m honestly not sure) about how he used an Amazon system that was .50 a minute or $30 for an hour of audio.

Comments are closed.