How Rev Wants To Make The World’s Conversations More Accessible To Everyone

Earlier today, I posted an interview with Wistia co-founder and chief technology officer Brendan Schwartz about his company’s mission to make people’s video more accessible to everyone. For Schwartz and team, better accessibility by way of transcripts and captions ultimately begets a better product for everyone, regardless of one’s abilities.

So it goes for Fernando Trueba and his own team at Rev.

In an interview with me earlier this month, Trueba, who serves as Rev’s chief marketing officer, explained the company’s mission as “[empowering] communicators globally.” Rev accomplishes this by leveraging artificial intelligence to provide people with “very accurate transcripts, captions, [and] subtitles” such that they are able to put out content and reach a global audience. Trueba’s tenure with Rev has been short thus far, having only been with them for about a year. He called his time an “amazing journey,” telling me the more he digs into the market and the company, the more excited he becomes about the “possibilities of empowering people and helping people with technology.” Despite technology’s capacity for goodness, however, Trueba was quick to caveat Rev is steadfastly committed to balancing automation vis-a-vis AI with human interaction. He told me the human element to Rev’s work plays a vital role not only in ensuring accuracy and fidelity, but also in terms of localization of different languages spoken around the world.

When asked about Rev’s raison d’être, Trueba reiterated the company’s overarching goal of empowering communication. More specifically, he pointed out that accurate transcripts and captions have relevance beyond accessibility. Like Schwartz said to me about Wistia, Trueba stated features such as captions go a long way in raising user engagement; Trueba said research has shown there’s 2% more engagement in videos where audiences can understand what’s being said of content. In general, Trueba said, people tend to feel more engaged with video when captions and subtitles are enabled because they facilitate communication and comprehension. He added people oftentimes watch videos in environments with various ambient sounds—for example, a student studying in class—so captions are a necessity.

“This goes beyond accessibility [in the disability sense],” Trueba said. “This is about people truly understanding what is being said and truly understanding the message for different reasons.”

Trueba triple-downed on Rev’s institutional ethos.

“What we’re trying to do is to enable content distributors, content creators, [and] storytellers to portray their message and to convey and connect with audiences at a deeper level,” he said. “We want to make sure that message is accurate, exactly as they planned it to be, because sometimes things can get lost in translation and things can get lost in technology. One thing that is intended to be said in one way could be understood in a different way if the means you’re using to communicate are not as accurate as they can be. That’s what moves us every day.”

For Rachel Kolb, using captions to communicate is everything.

Kolb, a disability advocate and fellow at Harvard University, is Deaf and told me in an interview concurrent to Trueba’s that captions provided by companies like Rev “opens the world” to her. She was unequivocal in telling me she cannot access spoken content without captions, adding they allow her to “get into spaces and conversations” that she otherwise would be excluded from. In other words, good captions equal inclusion.

A vital component of inclusion in Kolb’s case is accuracy; she needs translation to be correct. To that end, Trueba said Rev strives hard to ensure the “highest possible accuracy” of its transcripts and captions. To do so, the company cares a lot about the quality of the source data and from where it originates. He said humans are infallible and are prone to mistakes, and AI is similarly error-prone, so there has to be careful balancing of the two. Automatic speech recognition, or ASR, has become quite adept in recent years; Rev uses “thousands” of human proofreaders to complement the technical arm. Put together, the ASR and humans work “co-actively together” to produce transcripts and captions that are “as perfect as they can be.” Trueba said this relationship effectively creates a virtuous circle: because the output is virtually pure, that means the input being fed into Rev’s algorithm is virtually pure to the point that the algorithm itself indeed becomes “better and better over time.”

For her part, Kolb explained it’s been her experience that watching videos with bad captions give her headaches because she doesn’t understand what’s being said. As such, accuracy is extremely important to her and she commended Rev on its work in this realm. She said it’s never 100% perfect, but accurate enough that she’s able to comprehend what’s being communicated. In the last few years, AI tech has vastly improved that it’s become “more useably accurate” to Kolb. She added captions on social media platforms such as Instagram and YouTube to be “90–95% accurate” although she admitted not knowing what speech-to-text engines Meta and Google, respectively, use in creating captions.

Trueba said while captions and the like are obviously important for disabled people like Kolb, the reality is they are “absolutely” appreciated by people who are able-bodied. He cited accurate transcripts being used by lawyers for depositions or by journalists for interviews. Especially in the legal world, accuracy is of paramount importance because “getting a word wrong or getting context wrong could put someone in jail.” Thus, that transcripts be certifiably as accurate as possible matters a lot in certain industries. As to feedback, Trueba said Rev is a consumer-focused organization and works hard to “stay close” to its customers so it listens to them and hears what their needs are. He said the biggest concern for Rev’s users is accuracy, as people want to “spend less time correcting transcripts.” More high quality transcripts begets a better user experience, Trueba told me. He also noted customers want different ways to capture and edit audio, so a mobile app is important because people aren’t always sitting at a desk for hours on end every day.

Kolb again praised Rev for its accuracy.

“I’ve been happy,” she said. “The accuracy is a value of the company and I appreciate how this conversation is taking place about how captions are important for accessibility. That will always be my first priority.”

Looking towards the future, Trueba said his foresees technology becoming ever more collaborative insofar as he believes humans and AI will work hand-in-hand because it prevents bias. If, for instance, a story is told and a transcript generated, the text form retains all context of the conversation without interjecting a third party’s biases. This means the story can be presented to audiences in the creator’s vision, as they intended it. Overall, the work in captioning and transcripts is invaluable because, when all is said and done, the information exists as a form of communication. Rev, Trueba said, wants to be the supplier of that information in order to help people “learn faster and better, understand people better, and to improve the way we work.”

Kolb wishes for a world where she needn’t have to ask for access.

“My straight-up answer is I would love to see captions and transcripts everywhere. I would love to never have to make requests for them again,” she said. “We’re not quite there yet, but we could be. I think once we see more people take this up and see the benefits of captions and transcripts and just written tech for all kinds of applications, including accessibility but beyond accessibility, we could get to a point where I don’t have to go to the webpage for a podcast to be like, ‘Oh, I’m interested in that conversation, but they don’t have a transcript.’”