Listen to this article:
In spring 2016, the Danish digital magazine Zetland prepared to launch its daily news operation. The effort was crowdfunded, and the team at Zetland asked their supporters what they wanted from the publication. One thing kept coming up: audio.
But Zetland hadn’t raised money for an audio news operation. “We all come from a text background,” says co-founder Hakon Mosbech. “We just thought text would be the ideal medium to convey what we wanted to do,” which was to offer a few pieces of news and analysis each day.
After months of persistent requests from readers, Zetland “decided to do the most simple version,” Mosbech says: They had their reporters read stories into a microphone. The first audio article went online in fall 2016. “The user experience was definitely really clunky,” Mosbech says of the audio player embedded on the website. But people listened.
Narrated articles may not offer the most robust listening experience, but they do offer one that’s conve- nient for listeners
Soon, Zetland offered two audio stories a week. People listened to those, too. By 2017, Zetland reporters were narrating every story they wrote.
Now, when the magazine’s 14,000 subscribers open the Zetland app, they get a sort of playlist for each day, starting with a conversational podcast and moving into narrated articles. “In a way, it’s very old-fashioned the way members use us,” Mosbech says, comparing it to terrestrial radio. “They open the app in the morning and they just press ‘play’” and listen until they finish their commutes.
Zetland’s audio articles feature one voice, no outside sound, and no musical scoring — an aesthetic that’s closer to audiobooks than podcasts. In an age when many podcasts offer dramatic storytelling and sophisticated audio, narrated articles may not offer the most robust listening experience, but they do offer one that’s convenient for listeners and for the magazine: The audio articles serve people who are too busy to read but want to know what’s on Zetland that day, and they give Zetland access to people who prefer listening without requiring the publication to invest in the equipment and training needed to make a “This American Life”-style show.
Mosbech doesn’t see the audio articles as a replacement for podcasts. And he doesn’t see them as a replacement for text. Rather, they complement both platforms and give the audience something they want. And Zetland has found that the listening audience is a loyal audience.
Prior to the audio experiment, departing members said they didn’t read enough stories to stay subscribed. But “when people start using us through audio, they use us more and they use us in a more stable way,” Mosbech says. The average completion rate for an audio story is 90% — enviably high for anyone in a newsroom who has watched on a Chartbeat analytics dashboard as users abandon text stories a few paragraphs in. And because users hear a piece as part of a stream rather than clicking on a headline, members often listen to stories they wouldn’t otherwise read. This has remade Zetland. Other publications are poised to be remade, too. The success of podcasts has shown that audio journalism, much of it delivered digitally, has a sizable audience, editorial prestige (the first audio Pulitzer will be awarded this year), and — through podcast advertising or subscriptions to audio apps — a potential revenue stream. Narrated articles, which exist in between podcasts and text, can achieve these ends, too. The qualities that make audio engaging and valuable are present in both podcasts and narration. Harvard Business Review (HBR), The Economist, The New Yorker, and The Atlantic are among the magazines now offering an increasing number of narrated articles. Some make the narration themselves, and others turn to partnerships with news narration apps, which offer not only a share of the subscription fees the apps collect, but exposure to a new audience.
In the U.S. alone, the number of people who listen to spoken-word audio jumped by one fifth in the last five years; it’s now a daily habit for 121 million people (music listening, meanwhile, has declined 5%). That’s still far below the number of people who read text online each day, but publishers offering narrated articles find that listeners value depth with audio journalism, and, as they do at Zetland, will listen for longer than they read.
These listeners have largely moved beyond the perception — once often lobbed at audiobooks —that hearing text read aloud isn’t as effective a way of processing information as reading, and research offers promising support to back them up. For those not served by a publisher or an app, the technology for turning text into computerized speech is becoming more advanced and more ubiquitous; if audio isn’t available for a story, a smartphone app can create it, with increasingly human-like voices.
“Audio has kind of quietly emerged … as kind of the preferred way for people to consume content,” says Jim Bodor, the managing director of digital product strategy for HBR. “If audio does become, maybe not the dominant currency of the internet, but if it becomes an even more significant piece, then we have to be prepared for that.”
HBR is preparing for that future by partnering with Noa (an acronym for News Over Audio), a Dublin-based app that offers curated playlists of narrated articles from a variety of publishers. HBR began working with Noa in 2019, and as of January 2020, Noa had converted 32 HBR articles to human narration. Users clicked play on those stories 32,000 times. A quarter of the listeners stayed tuned for, on average, 90% or more of the piece. This is in line with other content on Noa. CEO Gareth Hickey says the average completion rate for stories in his app is 78%.
Two other apps — U.S.-based Audm and U.K.-based Curio — offer a similar service as Noa. And they operate in much the same way: They hire professional human narrators to turn text articles into audio, which they offer in a feed to paying subscribers. In exchange for giving the apps their articles, the publishers get a cut of the money and the ability to embed the audio alongside the text of the original piece (the apps’ founders declined to detail the revenue-sharing agreements).
Without offering in-depth details, the apps all say growth has been steady. On average, Curio has grown in users and revenue by 30% a month since launching in 2017, says founder Govind Balakrishnan, a former senior strategy manager with the BBC. Audm co-founder Christian Brink describes the 4-year-old app’s growth as consistent “since day one.” All three apps have attracted VC funding in addition to subscribers.
Beyond the fact that users are listening, publishers can find value in what users are listening to: deeply reported, context-rich stories
The apps pitch themselves to users similarly, too, taking advantage of the fact that audio’s increased popularity is tied to how convenient it is to listen while doing other things. HBR partnered with Noa after learning that they were missing out on these busy would-be readers. They learned that HBR users were not only saving articles to the app Pocket to read later, but were relying on Pocket’s automated text-to-speech feature to listen to the pieces being read by a synthetic voice. “If you don’t have the time to sit down and read a feature article in the latest issue of the magazine, we’re trying to deliver you something that makes it easier to get that,” says Maureen Hoch, the editor of HBR.org. “That’s very valuable for our most loyal audience members.”
And the proposition of finding new loyal audience members can be worth more to publishers than a share of subscription revenue. “The prerequisite to selling subscriptions is now engagement,” Hickey says. Besides offering long engagement times, Noa also gives publishers exposure to a user base that’s 65% aged 35 or younger. This makes for an attractive proposition. “The lead time to forming partnerships initially was many, many months,” Hickey says. “In 2018, it was a couple of weeks.”
Any publication could start having its reporters read stories, like Zetland did, but this takes training for the reporters (which Zetland offered to its team when audio stories took off), can be expensive to do on a large scale, and it doesn’t put the narration in front of new users. For publishers without Zetland’s base of paying audiophile members, or without The New York Times’s budget for making audio, an app partnership can be a shortcut to finding listeners.
Even publications that have been making audio for years have partnered with one or more of the narration apps, in part to find new audiences. The Economist, which is available on Noa and Curio, has made podcasts since 2006 and has offered subscribers a narrated audio edition of each issue since 2007. Like with Zetland’s narration, The Economist’s audio edition is a way of maintaining its subscription base. “We are a premium-priced publication and people feel guilty if they don’t read many stories,” says Economist deputy editor Tom Standage, who oversees audio strategy. “Our evidence suggests that the audio edition is a very effective retention tool; once you come to rely on it, you won’t unsubscribe.”
As popular as the audio edition is for subscribers, it remains just that—for subscribers. The app partnerships give The Economist a way to find new listeners (and they participate in revenue sharing, Standage says). “Partnerships like this give us an opportunity to expose new — and younger, more female, and more diverse — audiences to our brand,” Standage says. “The same is true of podcasts. We make money from podcasts, but strategically their aim is to reach potential future subscribers. Any subscribers converted from the apps may find the full audio edition is the best way to get their money’s worth.”
Beyond the fact that users are listening, publishers can find value in what users are listening to: deeply reported, context-rich stories. “I think the distinction between podcasts and longform audio articles and e-books is blurring, and that is likely to help preserve in-depth and investigative reporting to some extent,” Standage says, noting that podcasts have helped to give investigative stories new audiences.
The factors that make shows like “Serial” or the Center for Investigative Reporting’s “Reveal” podcast work so well also work for narrated articles. Listening sessions aren’t the same as glances at a smartphone. A story (or a playlist of stories) needs to last through a commute or workout, and it needs to engage the listener for the entire time. The format lends itself, therefore, to longer, more engaging stories. And even though narration can be produced more quickly than a podcast mix, audio is more valuable when it doesn’t become out-of-date quickly, so apps tend to feature stories that aren’t likely to expire shortly after they’re recorded.
When he worked at the BBC, Balakrishnan had seen in-depth reporting — one piece in particular that he recalls is a dispatch from a forward operating base in Afghanistan — languish online. “I always felt that there’s so much of journalism that gets buried into the daily news agenda. And this is timeless, well-researched, topical, deep content,” he says. Likewise, Christian Brink, the co-founder of Audm, says he built his app to bring longform journalism to more people.The apps often feature stories that don’t typically go viral on Twitter—the entire feature well from The New Yorker, say, or a long examination of John Updike’s oeuvre in the London Review of Books, all of which can be found in Audm. Even the more of-the-moment stories in the apps offer context and analysis rather than news updates. In early January, as tensions between the U.S. and Iran escalated and impeachment ground on, Curio featured analysis on Iran from The Financial Times, and Noa had a playlist of articles on Iran that offered background and context.
All of the apps, too, play the next story when one finishes, like on radio. If a listener avoids starting their commute by listening to a piece on climate change, for instance, they may very well end up hearing that story anyway, if they listen long enough. When I encountered the Updike piece in Audm’s feed, I added it to my queue for listening. I’d saved it to Pocket the week before, but kept putting off reading it, as some other task inevitably got in the way. Finally, while cooking dinner one night, I put it on. At first, I needed to rewind a few times to catch phrases I missed as my mind drifted or I checked a recipe. But soon, I was engaged in the piece.
In a piece written for the ear, information is presented in a way that acknowledges the audience can’t glance at a previous paragraph if they feel lost
A few weeks later, as I got ready to exercise, I noticed an essay from BuzzFeed in the Audm feed that had been sitting in an open browser tab on my computer for days. I ended up jogging a few extra minutes so I could finish listening. By that point, narrated articles had become part of my routine. I saved pieces from publications I already subscribed to and I discovered a few new writers as well. I was surprised how quickly I became inured to the bare sound of a single person reading, especially given that I work mostly in audio journalism and I spend part of every day adjusting levels, trimming sound bites, and otherwise using sounds beyond narration to make audio compelling.
Naturally, text-to-audio has critics, or at least skeptics. A single-narrator reading a piece meant for print, “is not necessarily using audio to the best end,” says Ma’ayan Plaut, a content strategist and podcast librarian for the podcast platform RadioPublic. The sound design that accompanies a documentary-style podcast or a radio news report is a craft that, like writing, can be appreciated on its own terms, but is also meant to help people understand a story and pay attention to it. Consider a 2018 NPR story about former coal miners with black lung, in which an affected miner struggles to keep his breath as he talks. The sound of his voice conveys information just as much as the words he says. With narration, this is lost. It seems counterintuitive, then, that this less-efficient way of storytelling has caught on as a way for busy people to consume journalism.
Further, not every story is written in a way that’s conducive to narration. In a piece written for the ear, sentences tend to be shorter, concepts are explained more conversationally, and information is presented in a way that acknowledges the audience can’t glance at a previous paragraph if they feel lost.
Mosbech believes one reason why Zetland’s audio experiment has worked so well is because the writers work to maintain a conversational style in their prose. At HBR, Hoch says stories that rely on charts aren’t prime candidates to appear in Noa. And Brink says Audm tends to avoid articles that are about niche topics or that feature long excerpts of difficult-to-parse language, like quotes from ancient poetry.
But with audio, convenience can trump the downsides. Besides content, the apps can also offer a level of curation that can be difficult to get with podcasts. The mix of publishers in Audm, the algorithm-driven recommendations in Curio, and Noa’s playlists based on single topics would take a lot of searching and user-level curation to replicate in a podcast app.
If users are increasingly preferring to listen to someone read rather than reading themselves, it may be good for the bottom line for apps and newsrooms, but is it good for the users? Is listening as effective as reading? Early research indicates it could be.
“The way the brain reacts is remarkably similar when you listen to something and when you read something,” says Fatma Deniz, a postdoctoral researcher in neuroscience at the University of California, Berkeley. Deniz and her team recently published the results of a study in which they placed nine subjects in MRI machines and monitored their brain activity as they listened to stories from “The Moth Radio Hour” — a public radio program that features people telling autobiographical stories live on stage — and as they read transcripts of those same stories. The scans showed similar brain activity whether subjects listened or read. Deniz says this indicates people may be able to process concepts comparably whether they listen or read.
Daniel Willingham, a psychology professor at the University of Virginia who was not involved in the Berkeley study, has written on the differences and similarities between listening and reading. Reading is relatively new to humans compared with oral communication. “Evolutionary biologists would say there hasn’t been enough time for the brain to evolve specialized adaptations for you to read, and so you’re using parts of the brain that evolved to do other things and then applying them to the task of reading,” Willingham says. “And in the case of comprehension, the process that you’re using is oral language comprehension.”
The key to understanding something read aloud, Willingham says, is prosody — the way words are said, the rhythm of speech, the emphasis and adjustment of tone. The sound of a voice carries tremendous meaning. Willingham cites sarcasm as an example. To say something sarcastically is easy — the speaker’s voice changes slightly. To write something sarcastically, however, invites misinterpretation.
When we read, Willingham says, we interpret words and phrases as if we’re hearing them. We add the vocal embellishments that a person telling a story on “The Moth” adds themselves. We even assign voices to quotes — if we know what the person sounds like (for instance, the president), we imagine the words in their voice. If we don’t know their voice, we invent one. A good narrator knows to do this, either because, as with Zetland, they’ve written the story and know how to convey the proper tone, or, as with the apps, they’re professionals who make a living interpreting text for the ear.
For example, in a narrated article on Audm, Noa, or Curio, the narrators do not say “quote” before reading a quote. Instead, they adjust the pitch of their voice. They may slow slightly if presenting a dense paragraph, or add emphasis to make certain points clearer. They will pitch their voice when they read a mid-sentence aside — like this one — and add a brief pause before and after.
This quality of narration isn’t yet matched by synthetic speech, which is one reason why the apps continue to hire humans to narrate stories. As voice assistants on phones, smart speakers, connected cars, and automated call centers become more common, however, we become increasingly used to hearing computerized voices, while the technology that powers them continues to improve. If we get used to having a conversation with computers, it seems inevitable that we could get used to hearing stories from them, especially if there are financial incentives for the quality of voices to improve.
Right now, the incentives to build an app around synthetic narration of news articles aren’t quite there. Play.ht founder Hammad Syed says his app grew out of a web browser extension that turned articles into automated speech. Unable to make it sustainable, Syed converted Play.ht to a service for publishers, encouraging bloggers, and other writers to pay for Play.ht to create narration or podcasts of their articles. “A lot of people engage with the audio, but almost half of them, they stopped before even completing 25%,” Syed says. When Syed asked users why they stopped listening, the most common response was that the speech sounded too mechanical.
As publishers think about where their partnerships take their stories, they’re also thinking about how these stories are written
To overcome this, Play.ht publishers can now make small adjustments to their synthetic readers to more accurately mimic the prosody of human speech. It’s not entirely clear whether this is enough to bring more users to Play.ht, but this is a recent advancement, and it’s likely the first of many improvements that will bring computer voices closer to human voices. “While it’s not for everybody right now, it’s only going to continue to be better,” says Josh Beckman, the founder of Narro, an app that lets its 300 users turn text articles into machine-narrated podcast episodes. Indeed, some apps, including Pocket and the recently launched news curation service Scroll, maintain their text-to-speech offerings for articles.
Automated speech reaching human-level quality would present a challenge to apps, as anything could then be turned into audio without any intermediary. Currently, though, the main challenge seems to be standing out from the crowd. The narrated article landscape is like a smaller version of the streaming video world. The three listening apps offer different content, just as Netflix and Hulu offer different shows. A user could pay for all of them, or pick one based on its publishing partnerships, its curation, and other features. It’s these features that the apps are using to distinguish themselves. Bodor says one reason HBR was excited to partner with Noa was the company’s potential to work with Land Rover and Jaguar to get their audio into connected cars. Balikrishnan says he’s thinking about location-aware content suggestions, with articles rising to the top of a user’s queue based on where they are.
As publishers think about where their partnerships take their stories, they’re also thinking about how these stories are written. At Zetland, Mosbech says the writing has “gotten a bit shorter, in terms of sentences and maybe also general length of articles,” in response to the success of narration. And 32 articles into its partnership with Noa, HBR is considering the role audio will play in its future. Hoch wonders if audio considerations might come earlier in the editorial process, rather than at the end when pieces are sent to Noa for narration. “If we get down to what is the point of all this: We are craving information, we are craving learning, we are craving stories being told to us,” Plaut says. “And the means by which it gets to us, I want to think, is more personal preference than anything else.” Or, as Mosbech puts it, “You should always be able to both read and listen.”