Jeff Kofman, is the Founder & CEO of Trint, after a 30-year profession with ABC, CBS and CBC News, Jeff got bored with hitting the wall of manual transcription and watching every story unnecessarily stop in its tracks. In 2014, Jeff and a team of developers leveraged AI to do the heavy lifting, and Trint was born.
Today Trint is an AI-powered SaaS platform that goes beyond transcription to spice up every stage of the content creation workflow.
From transcription to editorial tools, real time collaboration to export and publishing – making every step easier for newsrooms, podcasters, local businesses and global organizations to share stories faster and easier than ever.
You had a distinguished 30-year profession with ABC, CBS and CBC News, what were a few of the issues that you just encountered with traditional manual transcription?
I lived the pain of manual transcription day-after-day as a journalist. Long before I had the thought of inventing and constructing Trint, I wondered why there wasn’t a greater way.
Manual transcription was at all times the bottleneck in my workflow as a TV reporter. I’d do my interviews, hearken to a news conference, read my research, have a look at my footage and THEN… my producer and I’d disappear into the Black Hole of Transcription.
I can’t write my TV news story until I even have precise transcripts of the quotes or soundbites I need to make use of. I want to know what they said and the way long that soundbite runs. That meant sitting in a screening room or at our desks with headphones on, hitting PLAY then PAUSE. Then type some words. Then PLAY. PAUSE. And repeat. It could take hours. So tedious. So essential.
Trint launched in 2014, are you able to discuss how the thought was born?
I never imagined I’d be a tech guy. It was never in my life plan. It happened by probability.
I had an informal conversation with some software developers who had done some rudimentary experiments with audio and text (not transcription) in 2013.
I innocently asked: why can’t I take advantage of automated speech-to-text to transcribe my interviews.
I remember certainly one of the blokes asking me: why would you should do this?
I answered: because manual transcription is the pain point in my work as a reporter, I detest it.
We kept in contact and did some experiments. It quickly became clear that we had invented the longer term. I left my job as London Correspondent for ABC News a 12 months later and we began constructing Trint.
What were a few of the challenges of launching a transcription service in those early days?
Automated transcription is a discrete problem. Individuals who don’t live the workflow of reporters and content creators do not know how they create stories. I remember meeting some very wealthy angel investors within the early days and so they just couldn’t grasp why reporters like me need transcripts. It took a whole lot of explaining to get them to know how a reporter works.
I feel that’s easier today. We’re all content creators.
What are the several machine learning algorithms which might be currently used at Trint?
We’ve a brilliant smart bunch of engineers and data scientists which might be at all times tinkering with whatever they will be hands on with and wherever their imaginations can take them. As you’ll understand, our focus is on how automated transcription can speed up workflows for our media customers, which implies we at all times work around speech, speakers, languages and acoustics. NLP and speech processing algorithms are a part of our day-to-day, but we’ll investigate any creative ways to make use of AI to assist journalists extract information from videos, audios and pictures. Wealthy transcription lets us give more context to their content, makes all of it more searchable, and ultimately lets them find the moments that basically matter and get them out to their audiences as quickly as possible.
What languages are currently offered, and are there any differences in transcription quality between the several languages?
We provide about 45 languages which you could transcribe and are at all times adding more. Some are in “beta” and others which might be quite a bit more mature, which depends upon the scale of coaching data sets that help construct the models. We always measure the accuracy of our models for every language to always develop our models and to enhance their performance.
We’re at all times what latest models have gotten available to see if we are able to bring them into our secure ASR processing environment.
Nevertheless it’s not only in regards to the languages we transcribe – our customers may also have that transcript translated into almost any language.
Outside of transcription, Trint is an AI-powered SaaS platform that’s designed to enhance content creation workflow, are you able to discuss a few of the other tools which might be offered?
Although at the guts of Trint is our AI-powered transcription, what we obsess about is why those transcriptions are useful to our users, and the way we may also help them get value as quickly and simply as possible. Which means having a deep understanding of their workflows, so we are able to try to make every step as seamless as possible.
Ultimately, we would like them to capture any press conference, interview or event anywhere, at any time, in any language and make use of it because it happens. Which means making it easy for them or their team to confirm and use the live transcription because it happens – verifying, sharing and translating key quotes seconds after they’re spoken.
Our mobile app signifies that can occur even in the event you only have a phone on you, and ensures every part is securely transmitted to your team even when the connection is patchy.
Our Story Builder is designed to let you discover the important thing moments in all of your content and switch them right into a latest narrative that will be exported to other key tools in your content production workflow. Whether that’s a rough cut for video editing, a podcast transcript or an article. If it is advisable to use the text of the audio as captions, our collaborative editor may also help there too.
You furthermore may have a podcast that you just personally host called StoryTech, which looks at how technology shapes stories. Could you elaborate on what this podcast is, what listeners should expect, and why they need to tune in?
StoryTech is de facto the intersection of my two careers: a reporter and a tech inventor. It looks at how technology and innovation shape the best way stories are told.
The early episodes have a look at how CGI was used to bring down the ice wall in Game of Thrones, and the way the invention of the 35mm Leica camera within the Twenties led to the spread of photojournalism and the creation of LIFE magazine.
I’m fascinated by the impact of innovation on storytelling. That’s what StoryTech is about.
What’s your vision for the longer term of Trint?
That’s the challenge every innovator is wrestling with today. How does the fast pace of innovation open opportunities for my product?
Our customers desire a product that creates easy, intuitive efficiencies that fit seamlessly into their workflow. Which means going far beyond transcription.
Trint will leverage AI to do things that were unimaginable just just a few years ago: discover voices, faces, sentiment, context, facts and falsehoods. This can occur in any language – translating from that language because it’s spoken. The bottom line is to do that and so rather more in a way that integrates into other products to create one painless workflow.
I don’t see Trint replacing reporters, writers and content creators. It’s about liberating them from the drudgery of their work and allowing them to focus their time on creativity. It’s exciting to try to assume the longer term. I’m not gonna lie: it’s also daunting.
