As someone who records a lot of interviews for my day job, I’ve used numerous transcription services over the years. I’ve even paid for transcriptions from human transcribers, such as the excellent Danica from Upwork.
Whilst nothing can match the work of a professional transcriber such as Danica, online transcription tools are getting closer. Here, I compare four tools, each of which has its merits:
What to look for in a transcription service
There are five things that I look for in an online transcription tool:
- Automatic transcription… some transcription services only offer you an interface for you to type in the transcription, or expensive transcriptions by humans
- Understandable results – not 100% accuracy, but I do want to read through the results and for them to make sense
- Competitive price… ideally with a free option
- Intuitive controls and interface
- No horrible nags to upgrade or hidden costs
If you follow those five principles, you’re left with the four online services I’ve listed above. (If I’ve missed one that you love, please let me know in the comments and I’ll update this round-up.)
|Otter||Free up to 600 minutes per month||5/5|
|Temi||10c per minute||5/5|
|Transcribe||$20/yr plus 10c per minute||2/5|
|TranscribeMe||8p per minute||4/5|
Best free online transcription service: Otter Voice Notes
When it comes to free automatic transcription, nothing comes close to Otter Voice Notes. You get 600 minutes per month for free, which is a staggering amount – even if it was 60 minutes it would still be exceptional.
Despite this, Otter comes with some excellent features. If it recognises that a speaker is the same person that it’s previously transcribed, and where you have already supplied a name, then it will automatically recognise them and label the text appropriately.
Its automatic recognition is very nearly as good as Temi’s, with particular strengths around proper nouns: it correctly recognised “Moscow Bauhaus”, for instance, which others struggled with. You should always get results that you can read and understand on first pass.
Where it falls behind Temi is its transcription interface. There’s no highlighting of dubious words and your audio playback options are restricted to 0.25x jumps (eg from 0.75x to 1x to 1.25x) where Temi offers 0.1x increments.
However, if you’re willing to pay for Premium ($10 per month) then you’ll get more granular playback options along with custom vocabulary, email support and bulk export. Oh, and 6,000 minutes rather than 600.
That’s still great value compared to rivals and their typical 10-cent-per-minute charges, but you’re getting so much for free – including the ability to export individual transcripts and even share active pages with colleagues – that it’s a wonder people don’t stick with the freebie.
Most disappointing online transcription tool: Transcribe
Transcribe is in pole position when it comes to online transcription. It has the best name and it’s been established for many years. Sadly, it produced the worst automatic transcription results.
In some ways, age counts against it. The interface looks like it was designed a decade ago, and the automatic transcription feels like an add-on rather than an integral part of its service.
Instead, Transcribe appears more focused on the traditional transcription market: that is, a “pro” transcriber sitting at a desk with a foot pedal to control the playback while manually transcribing. And unlike Otter, which relies on mouse commands, it offers numerous keyboard shortcuts so that you need never lift your fingers from the board.
So as a manual transcription service it clearly has strengths. Where it falls behind rivals is its automatic transcription, with poor accuracy compared to Otter, Temi and TranscribeMe. On too many occasions, entire sentences were a nonsensical mess.
Nor does it tie the audio to the words – you even have to manually upload the audio file again if you want to listen along.
Add the fact that this is the most expensive service on test – $20 for an annual licence (which gives you access to its dictation engine, designed for manual transcription) plus 10 cents per minute of automatic transcription – and there’s little reason to choose it over rivals.
Best overall online transcription tool: Temi
Temi is arguably the most complete of the online transcription services on test. It will appeal to “traditional” transcribers thank to its high number of keyboard shortcuts, but it has a modern interface complete with trimmings: automatic speaker recognition, highlighted words if it isn’t confident, and support for video as well as audio files.
Most importantly, it’s also excellent at automatically recognising words. Even if it couldn’t get an entire sentence right every time, it came close and produced understandable results that were easy to edit on playback. Those confidence markers are also useful, but don’t assume that black text means it’s 100% right – I did find occasional errors.
There are a couple of absentee features, too. There’s no support for foot pedals and you can’t add custom words, which could prove a pain.
However, these are my only criticisms. I found Temi to be the most accurate automatic transcription on test (see my comments about accuracy below) and its interface – while not as intuitive as TranscribeMe’s – makes it a pleasure to use.
In short, even if you don’t want to spend money on transcription services, you should definitely give Temi a try. And the company makes this easy: your first file is free, no matter what the length and no credit card required. After that point, though, you’re paying 10 cents per minute.
Middle of the road: TranscribeMe
There’s an awful lot to like about TranscribeMe. Much like Temi, it offers a bunch of hotkeys for people who don’t want to lift their fingers from the keyboard, and it’s pay-per-play: 10 cents per minute, or 8p per minute if you choose the British pounds option.
And pay you must, because there’s no free trial here. That’s a shame, because it forces you to enter all your card details – there isn’t even the option to pay using PayPal – before you can start automatic transcribing.
The end results are pretty good; I’d put them on a par with Otter, but behind Temi. Like that service, it flags words it isn’t sure about, and generally it’s right that those are the areas to focus on.
One black mark is that, on playback, the audio quality drops compared to the original – all the services do this to some extent, I suspect as part of the algorithm to “sharpen” the sound, but I noticed it most on TranscribeMe.
The reason I still think it’s worth giving TranscribeMe a go are its superb editing tools, once you start going through text. For instance, if you’re correcting a word in the text then it doesn’t rush on to the next sentence – it stays with you until you’ve made your corrections.
If Temi had TranscribeMe’s interface then it would be an even clearer winner.
Analysis of results – how accurate are they?
You’ll notice I don’t quote “90% accuracy” or suchlike here. That’s because those metrics aren’t useful once you get going with genuine recordings. I could feed all four of these services a perfectly clear recording, from a speaker with great diction, using no unusual words, and they’d all score 100%. So what?
Instead – and I can’t emphasise this enough – you’re looking for a service that uses an algorithm that’s not only great at understanding words on their own, but also context. For instance, and I’ve no idea if this was luck or its algorithm, Temi recognised “VAT” in our interview with an accountancy firm whilst others went for “V 80”, “V he” or “VA T”.
If you’re desperate for figures, then I’d say that Temi is probably around 95%, TranscribeMe and Otter around 90% and Transcribe around 80%. But it depends on so many factors that I suggest you don’t get hung up on numbers. Give the services you like the look of a try and see for yourself.
How can you improve accuracy?
Accuracy comes down to the quality of the recording. While the built-in mics found in phones are decent, they’re no match for “proper” mics built into dedicated devices. It makes me look old-fashioned, but I swear by my trusty, chunky Tascam DR-40.
However, that costs £199, and has actually been replaced by the Tascam DR-40X for £170. If I was buying now I’d either go for that or the more compact Tascam DR-05X – which also happens to be cheaper at £99.
Not willing to buy dedicated hardware? Then use these tips:
- try to avoid noisy places for your interviews, including sounds you may not notice such as air conditioning
- always point the recording device towards the speaker… it matters less if you can’t hear your questions clearly
- advise your interviewee to avoid jargon if at all possible
- if there’s a way to politely ask your interviewee to either speak more clearly or, if they speak too fast, more slowly, then use all the diplomatic skills you have to do so
Thanks for reading and I hope you found this guide helpful. If you have a suggestion for another automatic transcription service to try then please let us know in the comments or send us a message via Twitter at @bigtechquestion.
Lead image by Tumisu from Pixabay.
READ THIS NEXT: What’s the best online translator?