Buried in my desk drawer is a scratched-up relic — a mini-cassette recorder that I used all the time as a young reporter to transcribe interviews. Now it looks like a discovery at an archaeological dig compared to my high-tech smart phone, which lets me record interviews for hours and share files instantly.
But even with this new technology, transcribing interviews from digital files hasn’t changed from the days of my ancient tape recorder. Even if I use my phone or a computer, I still have to hit play, type a snippet of what I hear, hit stop, rewind a little bit to my best guess of where I left off, and repeat the painful process all over again.
A new, fee-based service called Trint is trying to drastically streamline transcribing. And if you have quality audio, it does a pretty slick job.
“Getting the content out of recorded talk is still stuck in the 1960s or ’70s,” said Jeff Kofman, Trint’s CEO and co-founder who sat down for an interview with me via WebEx at Trint’s office in London.
In his former life as an award-winning foreign correspondent, Kofman was intimately familiar with the archaic, time-consuming problem of transcription. Working in television, Kofman often needed to grab just a few key soundbites out of a long interview, but it took precious time tracking down those quotes in his audio.
“In my 30-plus year career, all the technology has changed,” Kofman told me. “The whole workflow has been transformed in ways that we could never have dreamed in the 1980s — except this one part of the journalists’ workflow, which is how do we get the content out of our interviews?”
Trint tries to solve that problem by automatically generating a transcript of your recording. The transcript syncs with your audio. When you play the recording in your browser, you can follow the transcript “like karaoke,” Kofman says, and edit any transcription errors directly in the browser. No more ping-ponging between your audio player and Word document.
Here’s how it looks:
Proofreading an existing transcript can be a lot faster than transcribing from scratch. I used Trint to quickly find and snag key quotes from my interview with Kofman. I read the transcript and highlighted quotes that stood out for me. I listened to the recording to make sure the quotes were accurate. From there it was a simple matter of copying and pasting them into WordPress.
Trint — a combination of the words “transcription” and “interview” — offers various monthly plans but you can sign up for a free trial to test the techie waters. Plans start at $15 a month for an hour’s worth of recordings. If your files are longer you can continue to pay a quarter per minute as you go, and any unused minutes rollover to the next month. Kofman said this is a competitive price compared to professional transcription services.
“The whole point is to make it accessible,” Kofman said. “This is disruptive technology and it’s about making it easy to get a content and share.”
For my interview with Kofman, I didn’t use the audio from the WebEx recording — it was terrible. Kofman recorded himself on his iPhone and uploaded the video file to Trint, which can work with all kinds of video and audio files. He then shared the file with me in a process that’s like sharing a Dropbox folder.
For Trint to work well, clear audio is crucial.
“If you’re holding your iPhone 10 feet from the person giving the news conference, it’s not going to work,” Kofman said. “I mean, it’s really important to set expectations. Garbage in, garbage out. Good audio equals really good transcripts.”
Kofman said the most frequent complaint about Trint is that it fails to provide an accurate transcript. But in many cases, he said, the problem can be traced back to a poor recording.
Depending on the length of your file size, Trint can take a few minutes to prepare a file after you upload it. I recently uploaded a 19 minute, 23 second interview to Trint, and its transcript was ready to go within seven minutes.
I didn’t find many errors in Kofman’s transcript. (Ironically, Trint often failed to transcribe the word “Trint.”) You’ll have to keep an eye out for punctuation — Trint does periods but not commas and other types of punctuation. You’ll have to add those yourself.
When you work on a transcript, the text is front and center. To the right are your play and pause buttons and the video feed, if you’re transcribing video. Running across the bottom of the screen is a horizontal bar showing where you’re at in the recording. Click on the bar, and you can jump to that portion of the recording.
Trint offers several shortcuts:
- Hitting “Control” and the space bar plays or pauses the audio.
- “Control R” rewinds the recording five seconds.
- Outline some text and hit “Control H” to highlight a portion of your transcript. When you do that a yellow section appears on the bar at the bottom of the screen. Click on that yellow section and it will take you to that portion of the highlighted transcript. There’s also a “play highlights” button that will play audio only for those sections of the transcript.
- Outline some text and hit “Control S” to create a strikethrough and skip over a portion of the recording — if that’s your preference. The audio player will omit that section, and a grayed-out block appears on the recording bar.
You can set preferences in a menu in the bottom right corner of the screen:
In that same area is a “playback speed” option, which lets you slow down or speed up the recording.
In the upper right-hand corner are export options. These are handy. You can export your entire transcript in formats that include Microsoft Word or a zip file that combines an html file with the transcript of the audio recording. You can also email others to invite them to your Trint file.
Trint also gives you the option of exporting only your highlights, which I used to write this post. It created a Word document of everything I thought was potentially useful from my interview with Kofman, complete with time stamps. Nice.
Kofman said Trint takes security and privacy very seriously, but it’s important to keep in mind that it’s a cloud service.
“We’re very sensitive to that,” Kofman said. “We are encrypted end to end. Our goal is to get the International Standards Organization data security certification. It’s a one to two-year project. You don’t do those kind of things overnight. It’s a really, really laborious and time consuming and resource sucking project and it’s very much a priority.
“So you know, I’m very honest,” Kofman added. “If you’ve got the next Edward Snowden, I would be cautious.”
When I first tried Trint last year I thought it was buggy. When I tried revising a transcript, Trint screwed up my typing and introduced errors. I initially wrote it off as a glitch in a beta product and assumed the company would work it out.
What I didn’t realize is that Trint works, but some Chrome plugins interfere with it. When I emailed Trint last month to ask what was causing the typing bug, Kofman himself replied and said deactivating plugins for Evernote or Adobe usually does the trick. And it did. I also found going to “incognito mode” in Chrome resolves the glitch.
It might have been my Internet connection but sometimes when I moved the cursor to a spot in the transcript and hit play, Trint lagged for a while before playing the audio. Hitting “Control R” to rewind the audio five seconds works but the shortcut itself is a little clumsy — I kept having to look down at the keyboard to make sure I was hitting the right keys.
I often use another transcription tool called oTranscribe, by Elliot Bentley, which uses the “escape” key to play audio and pause and rewind five seconds. Hitting escape is more intuitive. As I was working on the interview of Kofman’s transcript, I also craved an annotation feature to jot notes to myself.
Kofman said annotations and the rewind shortcut are things the development team will be looking at as they work on Trint 2.0, which he said will be offering new collaborative features that will make it easier to share content.
“We’re developing what we call the Trint player, which will allow you to take an interview, a news conference, a lecture, clean it up and actually put it on your website so that it instantly becomes searchable,” Kofman said.
The Trint player is scheduled to be released sometime in mid-2017. It will allow readers to outline a soundbite and share that segment of transcript and audio to social media services such as Facebook and Twitter.
“Right now if you put a 20-minute interview on your website or in your archive, it’s dark data unless there’s a complete transcript,” Kofman said. “You don’t know what’s on it. What we do is shed light on that dark data because we actually allow you to find it and to access that content quickly and then put it on your website and make it SEO, search engine optimizable. Suddenly we can drive traffic to that news organization or a corporate Web site or education web site, which is really important.”
Other plans include one-click captioning, which will automatically bake the captions of your choice into a sharable video. This could be a huge time-saver as subtitles become so important in Facebook feeds.
“That’s a fundamental innovation of Trint,” Kofman said. “We allow you to highlight a moment and send that because we glue the audio to the text.”
What about a smartphone app? Kofman said he recognizes there’s demand for a mobile app but for now Trint will remain a browser-based tool as his team focuses on core features.
“A smartphone app is very much on our road map,” Kofman said. He’d like to see an app that would let people record an interview, then upload it directly to Trint from their device.
Critics might complain that Trint fails to provide transcripts that are 100 percent accurate. Trint appealed to me because no matter how accurate a transcription service claims to be, I still have to listen to it myself to make sure every quote is accurate before I use it in a news story or, in this case, a blog post. Trint simply makes that process easier.
And it really shines when you’re looking for something specific — that golden soundbite you vividly remember but can’t quite find in the recording. In Trint, it’s as simple as typing “Control F” in your Windows browser to find a keyword and quickly pinpoint it. This feature alone can be a real timesaver.
“I had a reporter with one of the NPR affiliates tell me she had six hours of interviews for a long feature she was doing,” Kofman recalled. “And she said normally that would be two days of transcription for her. She would just lock herself away and type away. She said she uploaded it to Trint and she got it back in less than an hour and she came in the next morning and she went through it. She knew her material. She was able to scan and search it and she said she had the entire story written before lunch.”
Not surprisingly, news organizations rank high among Trint’s major clients. Early corporate accounts included ESPN, Vice News, and the Canadian Broadcasting Corporation. “Fast Company came to do an article on us. They’re now one of our biggest users,” Kofman said.
But Trint’s appeal goes beyond journalists. Kofman was surprised by its growing popularity in academic circles — researchers and students who have to conduct hours of interviews and make sense of them.
The team teases me and calls me ‘Beta Tester No. 1.’ And it’s true. I have lived the problem.”
— Jeff Kofman
Trint is a small but growing company that’s expected to double by summer from 10 to 20 employees and open a new North American office. It’s received grants and investment money from the Google Digital News Initiative and the Knight Foundation’s Enterprise Fund. Cisco Systems, the company behind WebEx, sponsors Trint’s work space at IDEALondon, an innovation lab in Shoreditch in East London.
“We’re sponsored here in London, England by Cisco because Cisco sees huge application in the corporate world for corporate communications as the audio gets better,” Kofman said.
It’s no accident that Kofman, as a former journalist, immediately recognized how useful something like Trint could be when he met the original development team.
“The team teases me and calls me ‘Beta Tester No. 1,'” Kofman said. “And it’s true. I have lived the problem.”
After 30 years in the news business and learning how to adapt to rapidly changing circumstances in far-off places like Libya, Kofman is using those skills in the tech world as an entrepreneur, asking lots of questions, learning as he goes, working long hours — and having fun along the way.
After seeing what Trint, in its early form, could do, Kofman said he didn’t have much choice but seize a rare opportunity. In 2013, Kofman was still working for ABC News teaching a global journalism program for U.S. and Canadian students studying in London. A friend urged him attend Mozfest, a big media coding conference that happens in London each November.
“That’s where I met three brilliant developers who are the core of Trint’s tech team,” Kofman recalled. “They were demonstrating a transcription experiment that glued manual transcription to the original audio.”
Impressed with what he saw, Kofman asked Laurian Gridinoc, now Trint’s senior developer, if it could work with automated speech-to-text.
“Interesting idea,” Gridinoc replied. “We could try it.”
“I remember saying, ‘This is the future,'” Kofman said. “Either we team up and we make it happen or we’re going to be sitting in a coffee shop five or 10 years from now looking at someone doing something like that saying, ‘You know, I remember talking about that but we didn’t do it.’
“And I just thought, we’re going to do it.”