Transcribing audio — it couldn’t be easier!

If you record your teacher, lecturer or professor during lessons, then you’ll want a quick and easy way to transcribe what they have said. The same is true if you are, for instance, a journalist, and interview people.

Microsoft Word has a dictate feature. You open an audio file, turn on the recording on, press ‘Dictate’ and Word will transcribe it for you.

Alternatively there is a paid service called subeasy.ai, based in Hong Kong, where you can get three free tries.

At this point, I should mention that I am a university lecturer in scientific communication for non-native speakers. I have absolutely no connection with SubEasy — I simply think it is a really useful tool that you may not know about.

The ‘sub’ in ‘SubEasy’ does not refer to ‘subediting’ (meaning checking and correcting a text before publishing it), which was the reason I first clicked the link to the tool when I read it in AI Secret — an AI newsletter. I thought it would help in text editing!

It actually refers to subtitles. The founder, Terry Gu, explains where he got the idea from:

Last year, I got the amazing news that I was going to be a dad. My wife, who’s really into TV dramas — especially those focused on Japanese urban life and ancient Chinese romance — started watching a lot of shows while waiting for our baby to arrive. She couldn’t get enough from Netflix and even subscribed to a ‘local Netflix’. But she quickly ran into a problem: there were no translated subtitles for the shows she wanted to watch.

His solution was:

First, get the video and extract the audio. Then, transcribe the audio into subtitles with high accuracy. Finally, translate those subtitles and watch the video with the synced, translated subtitles.

For me this is a classic case of how an AI comes into being, and how it can also be used for tasks that the creator may never have even thought of when developing the tool.

In the world of education, while there may be occasions when you want to create subtitles for a presentation or a video abstract, you are much more likely to want to transcribe a recording that you made of a professor or lecturer talking. And this is where SubEasy comes into play.

As my test case, I used a recording (mp3) made for me by a female PhD student, Michelle, from the Philippines. She talks about Catholicism and Easter in the Philippines. She speaks reasonably slowly with a slight American accent.

The example below is a ‘Compare Documents’ of Word’s Dictate (original document) and SubEasy’s transcript (revised document). It shows you just how powerful SubEasy is compared to Word’s Dictate. In yellow I have highlighted the major mistakes Word mistakes. In green is the only mistake SubEasy made. It should read: … fasting, abstinence, pray, go to church …

The advantages of SubEasy are:

· It punctuates correctly.

· It recognizes where one sentence ends and the next begins.

· It spells ‘technical’ words correctly e.g. panata and not pinata.

· It differentiates between similar sounding words e.g. whole vs holy.

· You can generate a text summary of the audio file you have uploaded

· It is much quicker. Dictate requires the entire text to be dictated. With Subeasy you just upload the recording.

Another advantage is that you can press ‘Translate’ and it will translate into many different languages. The ‘cost’ of the translation depends on which option you choose. SubEasy offers a choice of Microsoft Word for free, SubEasy AI (costs you credits and gives ‘excellent’ quality), or SubEasy AI plus (which costs more credits and gives ‘superior’ quality).

The translation is done paragraph by paragraph, so you see a paragraph in the original English and then a paragraph in the target language. This makes it easy to check whether the translation is accurate. You can then export the file. The original file just shows the translation, without the original English.

I got SubEasy AI to translate what Michelle said into Italian (my second language). I then got Claude to check it. Claude made the following changes:

· three punctuation marks

· put panata into inverted commas (“panata”) to highlight that this word is foreign (it means ‘vow’ in Tagalog, the language spoken in the Philippines)

· it changed sicuro (which means ‘sure’) to ‘sicura’. In Italian an adjective ending in -o indicates a masculine subject, and -a a feminine subject. Michelle is a woman, therefore the correct word is sicura. In Italian, the name Michele (with one L) is a boy’s name — Michael.

So really, SubEasy had just made one mistake — the masculine instead of a feminine adjective. As a professional translator, I would say that that is an excellent result.

If I had translated Word’s original version using Word’s own translator, the result would have been a disaster given that Word’s version in English already contained a lot of mistakes. In fact the word ‘spray’ in Word’s version in English also appears in Word’s Italian version, thus making the sentence totally incomprehensible!

If you are into podcasts, SubEasy can also help you convert the audio into text, as explained in this blogpost. This is particularly useful if:

· your listening and concentrations skills are not sufficient to be able to follow a long podcast

· you would prefer to ‘read’ the podcast in your own language

· you want a chatbot to summarize and individual podcast or even a series of podcasts

· you want to a chatbot to compare what has been said about x in one podcast with what has been said about x and/or y in another podcast

Another use, suggested by the creators, is to use SubEasy to organize your interview or meeting notes. This is described in another of their blog posts. This feature is useful if

· as part of your research you carry out interviews

· you want to make notes of what was said during project meetings

If you have ever wondered why people pay for a service that they could get free elsewhere, I think the SubEasy example makes a compelling case.

Adrian Wallwork

· Since 2000, Adrian Wallwork has held more than 100 courses for doctoral students and researchers at universities primarily in Italy, but also recently in China, Qatar and Turkey.

· Author of textbooks for Springer Science, Oxford University Press, Cambridge University Press, BBC, De Agostini and Vallardi. His books for Springer Science on English for science communication are used by universities around the world and have to date been translated into Chinese, Korean, Japanese and Persian/Farsi.

· In February 2023 he began teaching PhD students and researchers how to improve their English using Google Translate and ChatGPT — with an overall increase in accuracy from about 75% to 95%.