This transcription tool could be a game changer for newsrooms
Trint won GEN’s Startups for News competition at the GEN Summit in Vienna in June 2016. This September, Trint officially launched. CEO & Co-Founder Jeff Kofman, Emmy-winning war correspondent and foreign correspondent who spent more than 30 years reporting for ABC News, CBS News and CBC News, answered a few questions for us…
GEN: Can you tell us a little about your launch? Are you excited?
Jeff: Excited? Trint began almost two years ago as a crazy idea that had the potential to transform the daily lives of journalists. It was incredibly ambitious to think that a little team (we are now eleven but originally we were three brilliant developers and me) could find a way to automate the transcription process. But we’ve done it! I am amazed and really proud.
The launch? When we opened for free testing last February so many people came at us so fast that we went viral. We weren’t ready, we had to close the site down and control access while scrambling to finish the build. Now we are ready. I think we are on our way to becoming a cult. We constantly hear users call Trint ‘magic’.
How do you think automation can help newsroom without discrediting the journalist’s work?
From the beginning we understood that the flaw in machine-generated transcription is that it makes errors: a technical term, an obscure name, or simply bad audio. An executive at CNN told us very early on something that has become our war cry:
“We want your project to succeed, but understand that we will never touch it until we know we are working with transcripts you can trust.”
Transcripts You Can Trust. That’s exactly what we deliver. Because the audio is glued to the text you can search the automated transcript, verify it and if there’s an error you can correct it. Journalists are telling us that are reducing hours of work to just minutes.
What is the next big innovation coming to video according to you?
I can tell you that one of the most recent innovations in video is posing a real challenge for news organisations: captioned or silent video. It’s hugely popular on social media and websites. But it’s not easy to produce: it requires transcripts, separate video, and specialized editing skills. With Trint we have simplified the workflow to one-click. We can literally give you burned in captions on your audio or video in just one-click. When I demonstrated One-Click Captions to a group of news executives in New York a few weeks ago they spontaneously burst out in applause.
Can transcription be used in parallel with translation in the near future? This could be a real game changer, especially for live videos…
That’s a question I am often asked. It’s already technically possible, the challenge is that you are running spoken word through two separate algorithms: one for transcription, the second for translation. The problem is that badly-recorded audio or heavy foreign accents mean that the mistakes of the transcription are amplified exponentially when they are passed through automated translation. That said, in certain circumstances — broadcast audio that is crisp and clear — it can work. It’s on our roadmap but it needs more research before we can fully develop it.
Newsrooms often have a lot of videos in their archives. Is a tool like Trint able to give those a second life thanks to transcription? How is that?
Absolutely. Those videos in archives are called ‘dark data’. Currently the only way to know what is on them is to have someone listen to them and manually transcribe them. That’s just not cost-effective in most scenarios. With Trint an old interview can be run through our system in minutes (we call it “trinted”) and then instantly searched for the necessary content. Once we build our full archive search tool you’ll be able to instantly look through thousands of Trints and see and hear the relevant sound. It’s a real game-changer.
There is a lot of speech-to-text software on the market. What makes Trint different?
Automated speech-to-text simply gives you an error-filled text file. It’s of little value to journalists because there is no easy way to identify where the text is correct and where it is wrong. By marrying the text to the original audio and allowing users to correct machine-generated errors, Trint opens speech-to-text to vast new markets. We make it easy to find content, instantly share it on social media or make it search on the Internet. Transcripts You Can Trust. It’s a whole new world for transcription.
What will be Trint Ltd’s next new feature? Can you give us any hint?
You want us to reveal state secrets? We recently won a grant from the Google Digital News Initiative to support our next phase of innovation. The team is working on it now. I can honestly tell you it is jaw-dropping. The few people who have seen it often simply say: “Wow!” Stick with us!