Do you have transcription tips?

November 3, 2019 3:37 PM

I need to transcribe 15 to 20 hours of interviews. Someone recommended Express Scribe software and the foot pedal, but I see that it gets pretty mixed reviews. Do you have any tips for transcribing efficiently? What has worked for you? I have a Mac, and I'm recording my interviews both on my iPhone and also with an Olympus recorder. I've found AI transcription software to be useless. And for various reasons, I'm not keen on paying someone else to do it.
posted by swheatie to Writing & Language (17 answers total)

Express Scribe is fine, and you can start with the free version, so I'd go with that. Foot pedals are a personal preference thing -- if you're practiced, you can definitely go faster, but it demands some coordination skills, and you may find the learning curve isn't worth it for your relatively short project. If you game with a pedal or play drums or otherwise are experienced with using your hands and feet simultaneously, you may prefer it. But you can use keyboard shortcuts to handle what you'd otherwise do with the pedal.

Whether or not you go with a foot pedal, make sure your ergonomic set-up is decent, and take frequent breaks to stretch. It's too easy to fall into the trap of "I'll take a break once this interview's done," I find, so set a timer to ensure you're resting and moving around frequently enough. It's very easy to get into the zone, and very easy to unconsciously keep your muscles tensed up to react in time to pause and rewind a fiddly bit, which means it's very easy to end up with arm, shoulder, back, or leg pain. I've definitely wrapped a long interview and suddenly realised that my leg's asleep or my hand's going numb, and it's a bad time, so don't do that.

The biggest favour you can do for yourself is to record your audio at the highest quality you can manage. If you're still interviewing, work on refining your technique. (Also, get room tone!) But if you're done the recording phase, and you're finding your quality isn't great, Audacity is pretty good and is free.

I find noise-cancelling headphones to be helpful -- you don't need a branded-for-transcription headset, but headphones are definitely preferable to speakers, and you want to have decent ones.
posted by halation at 4:01 PM on November 3, 2019 [2 favorites]

I have used a transcription pedal in the distant past, but just wanted to suggest potentially trying a different method of recording (I use Skype + Call Recorder, personally). I have transcribed literally hundreds of interviews with Otter, with a huge variety of accents and different qualities of phone connections, and it has worked very well, so I'm curious what issues you're having.

As an alternative to a foot pedal/software, you can also upload audio to Otranscribe. It's definitely an improvement from trying to type in a standard Word document, since it lets you use shortcuts to pause, rewind, etc.
posted by pinochiette at 4:13 PM on November 3, 2019

If you do want to try something that's not Express Scribe, though, F5 is great for Mac. Unfortunately, I don't think there's a free trial, though you can play with it a little to test it out (first 10 minutes of a file, I think). You could also give InqScribe a try. It's more intuitive than Express Scribe, and has a 14-day free trial, but it is less customisable.
posted by halation at 4:17 PM on November 3, 2019 [2 favorites]

I have a foot pedal and quite like it. I also have a folder in TextExpander that I activate when transcribing so that I can type shortcuts for common phrases or annotations.
posted by 10ch at 4:53 PM on November 3, 2019 [1 favorite]

Seconding using TextExpander, Keyboard Maestro, or similar macro program for typing shortcuts for common phrases. Just for names alone it can be a relief.
posted by Mo Nickels at 5:14 PM on November 3, 2019

Although you've found AI transcription software useless, it might be worth trying again: I've had a lot of success with poor-audio interview transcription using Otter.

If you haven't tried Otter, uploading your audio file via the web interface at Otter.ai would likely be better than using the app.
posted by anadem at 5:19 PM on November 3, 2019

I like listening to the interview I did on headphones while repeating the words myself, enunciating carefully, into voice dictation software, like google docs or apple's built-in software. That works better than just feeding an interview into the software, I've found.
posted by umbú at 6:37 PM on November 3, 2019

I've used Trint. And yes, while AI makes a lot of mistakes, it's still easier to start with the AI rough draft and then edit as you listen.
posted by RedEmma at 8:21 PM on November 3, 2019

I'm definitely happy with Otter.AI (referral link) and then doing a second pass to edit the errors - saves me heaps of time over transcribing from scratch.

But YMMV if you have a lot of background noise, lots of overlapping voices, or if you have other languages than English in the audio. (It does do well with a variety of accents, though). It's also not suitable if your transcription or audio needs to be strictly private. (I'm actually impressed with Otter's privacy policy and it's probably fine for most stuff, but e.g. if you have a legal requirement to keep the recording secure, e.g. healthcare stuff or whatever, it's probably not going to fly).
posted by lollusc at 10:47 PM on November 3, 2019

Decide before starting what quality of transcription you want. Just meaning? Every worc? Every um, ah? This should be coloured by what you want to do with it once you are done, so have a clear idea of that first.
posted by biffa at 12:05 AM on November 4, 2019 [1 favorite]

I've done transcription for money in the past, and what I did was opened up the sound file, then opened up Word and shrunk the window a little. Then I'd listen to the sound file, pause it, then click over to the Word document and type. Goes pretty fast once I get into a rhythm. This is using a Mac with an external keyboard and mouse hooked up, not sure I could do it as easily with a track pad.

You can get a cheap wireless Logitech keyboard and mouse, and hook it up to a laptop via USB and balance the keyboard on top of the laptop's keyboard and use the mouse on your desk or table in front of it. I find it so much easier to type on a regular keyboard than a laptop keyboard.

I've used an old school Dictaphone with a foot pedal, and one advantage of doing it onscreen is that I can see the second mark where I stopped, and easily click back a little in the timeline if I've missed something. If I get words that are unintelligible, I use a place maker word, like, WORD in caps, so I can keep going.

I just use Q. at the beginning of the interviewer's question, then A. for the interviewee's answer, while listing their names at the top of the document. You could also use initials if that works better for you.

Since these were both limited projects, about 10 interviews each (around 20 minutes average per interview), I didn't want to invest in special software and equipment, so that's what I did, and it worked fine for me.
posted by Marie Mon Dieu at 2:36 AM on November 4, 2019

I just bought a transcription pedal for a project with poor audio quality and foreign accents speaking about technical topics (tried Trint but this particular file was just too rough to get a good automated transcription).

I’ve used OTranscribe in the past but the pedal made an enormous difference. Took 20 minutes or so to get used to it but I would highly highly recommend if paying for outside transcription is off the table for you. I got a USB pedal that came with Express Scribe software.
posted by forkisbetter at 4:50 AM on November 4, 2019

If you change your mind about using a paid service, Rev is $1/ minute and have done good work for us.
posted by metasarah at 5:30 AM on November 4, 2019 [1 favorite]

If you use a paid service now or in the future, please select one that pays its contractors a reasonable wage for what is (as you have seen, if you've tried to do it) simultaneously highly skilled and highly tedious work, and a kind of work that requires the full attention of at least two senses, precluding all the usual distractions that make other kinds of office/editorial work tolerable. Any company that is A. not a single person and B. returns high-quality transcriptions with a speedy turnaround is using editors as well as transcribers, and both roles are frequently filled by contractors. that is a lot of people who are all getting tiny fractions of the advertised price per audio minute.

this is the same ethical issue that arises when hiring house cleaners. transcription is a shitty career, but a real one.

If you're positive this will be the only audio project you ever transcribe, yes, you can get by awkwardly switching windows and rewinding a lot. If there's a possibility of doing more interviews at some point, or if you value efficiency, I would absolutely buy a pedal to use with Express Scribe. You may find that by calculating your own hourly pay rate against the time saved by using a pedal, you aren't losing any money overall.

I should add that hourly contractors have to hit a 4 or 5 to 1 ratio (hours of transcription time to hours of audio) to make anything close to minimum wage, but if you take breaks now and then, if you have to pause to research terms and spellings, if the audio quality is imperfect, or if you need to proof your transcripts, it will be more like 6 to 1. so this is a project that (as you may already have calculated) you can expect to spend around 80 hours on, unless the speakers are very slow, clear, and take a lot of long pauses.
posted by queenofbithynia at 11:39 AM on November 4, 2019

You may have limited luck with AI transcription if you have multiple alto-or-higher voices. Ultimately, that was what drove me from AI+cleanup to Express Scribe+foot pedal doing it all manually. I transcribe a lot. Worth it the investment of time and money into the pedal, the AI never worked well.
posted by branca at 12:16 PM on November 4, 2019

Back when I did audio transcription, I used LibreOffice Writer because it learns your vocabulary as you type and autocompletes words. For highly technical content or complex place names/proper nouns, it was invaluable.

You can also copy/paste a corpus of text into a new document and it adds the vocabulary to the cache for that document. So you can delete that and start typing with preloaded autocomplete.
posted by overeducated_alligator at 2:18 PM on November 4, 2019

Otter.ai and a second pass is free and quick
posted by Andrew Galarneau at 12:38 PM on November 6, 2019