This essay first appeared on the Medium blog Generative AI in the Newsroom and is republished here with permission.
Journalists have been using audio AI tools for some time. Software like Otter.ai allows journalists to transcribe interviews easily, and Natural Readers can read an article aloud for those who prefer listening.
Recently, tech giants OpenAI and Meta have made major strides in voice and speech AI. Meta announced that their AI will respond in a voice that closely mimics celebrities and OpenAI rolled out a new API that allows users to speak with their models in real-time.
The need for improvement in radio and podcast news is clear — 50% of U.S. adults never get news from these mediums, with only 9% using them often. Adults aged 18–29 are the least likely to use audio news, with just 2% preferring radio.
With these advancements from OpenAI and Meta, newsrooms may adapt and build new features to enhance the reader and listener experience. This post explores how newsrooms may use news voice and speech AI technologies to attract more consumers and improve their experience.
Meta Celebrity Voice AI
Meta’s advanced AI featuring celebrity voices offers a unique, personalized way to interact with various forms of content. Interacting with this AI is straightforward: users simply choose from a selection of celebrity voices, and the AI generates spoken content in the chosen voice.
This AI could open new avenues for how news is delivered and consumed, offering a more personalized and engaging approach to news:
1. Personalized News Briefings
Newsrooms could develop a feature that allows users to select from a variety of celebrity voices for their daily news briefings. This could attract listeners who might prefer the familiar tone of a celebrity voice, adding entertainment to routine updates. The ability to choose between different voices could make news consumption feel less like reading a formal report and more like listening to a favorite podcast.
- Example: Imagine getting your morning news update from Judi Dench’s sophisticated narration or having John Cena recap last night’s sports highlights in his energetic style. This customization could appeal especially to younger audiences or those seeking a more engaging way to catch up on the day’s events.
2. AI-Generated Podcasts and Explain-It-Like-I’m-Five Segments
Creating podcasts or explainer segments where familiar voices tackle complex topics is another avenue for newsrooms to explore. AI-generated voices could break down difficult subjects like economic policy or climate science into digestible audio content, making it easier for listeners to understand and stay engaged.
- Example: A weekly political recap podcast, “hosted” by the AI voice of Awkwafina, could bring a humorous and relatable tone to otherwise dry policy discussions.
These AI voices could make listening to complex subjects feel more like chatting with a friend than sitting through a lecture, which could be a game-changer for news outlets looking to attract new listeners and maintain engagement over longer episodes.
While some newsrooms work hard to make their human-hosted podcasts engaging, with hosts using a conversational tone, human narrators have limits. They can only appeal to certain audiences.
AI-generated voices offer a way to broaden that appeal. By allowing listeners to choose from various voices, such as a favorite celebrity or a humorous tone, AI could engage those who don’t connect with traditional hosts. Combining human and AI voices could help news outlets reach more people and make complex topics more engaging for a wider audience.
By adding personality to real-time information delivery, Meta AI’s technology could offer a unique edge for newsrooms aiming to create memorable, interactive content. For a younger generation who seems hesitant to consume news in this medium, this could be a game changer.
Realtime API
OpenAI’s Realtime API offers exciting possibilities for media companies aiming to enhance how they deliver and interact with live content. This new feature enables AI models to process live audio streams in real-time and offer immediate responses.
It operates by receiving continuous audio input, which is instantly analyzed and converted into text or other outputs with minimal delay. This process happens as the audio flows, allowing the API to provide near-instant feedback without the need for batch processing or waiting for the stream to finish.
The API’s low latency ensures smooth, real-time interaction, making it adaptable to various live audio environments like broadcasts, phone calls, or events. By breaking down audio into manageable segments, it processes each piece as it arrives, delivering responsive and seamless output.
Here are some innovative ways newsrooms could leverage this technology:
1. Instantaneous Interactive Reporting
With the Realtime API, media companies could offer real-time Q&A sessions that let audiences interact directly with ongoing events. This capability could transform the way live coverage is delivered, enabling users to get immediate answers to their questions while following news developments.
- Example: Imagine a live stream covering a natural disaster or political event, where users could ask, “How many people have been evacuated so far?” or “What’s the latest update on the vote count in Georgia?” The Realtime API could pull relevant data and respond immediately, turning the coverage into an interactive experience. This level of engagement could make audiences feel more connected to the unfolding story, providing a sense of involvement that traditional broadcasts cannot match.
2. Adaptive News Summaries as Stories Develop
The Realtime API could also enable live-updating summaries of evolving news stories. Rather than waiting for complete updates, audiences could see summaries that adjust as new information arrives, keeping them up-to-date without needing to refresh the page or watch long broadcasts.
- Example: For instance, during a major breaking news event like an election or a court ruling, the Realtime API could continuously update a summary of key points, providing bite-sized, timely updates. This would allow media outlets to offer a dynamic “live blog” experience that is more efficient and accessible for readers who want the latest information without sifting through a full article.
This kind of real-time adaptability ensures that audiences are never behind the curve, making the news feel fresher and more immediate.
3. Real-Time Fact-Checking During Live Broadcasts
Newsrooms could use the Realtime API to integrate instant fact-checking into live coverage, a crucial tool for maintaining credibility in today’s environment. As reporters cover press conferences or live events, the API could flag inaccuracies in real-time, allowing journalists to correct the record almost immediately.
- Example: During a political debate, the Realtime API could instantly verify facts or identify discrepancies between a candidate’s statement and known data. This capability could be integrated directly into a broadcast, offering viewers real-time fact-checking graphics or updates. Such a tool would give newsrooms a significant edge in the race to provide accurate, up-to-the-minute coverage, reinforcing their role as trusted sources of information.
4. Enhanced Interview Assistance for On-the-Go Reporters
For journalists conducting live interviews or reporting from the field, the Realtime API could serve as an on-the-fly research assistant. This feature could suggest relevant questions or provide background information in real-time, enabling reporters to dive deeper into the conversation and adapt their approach as new information emerges.
- Example: During a live interview, a journalist might need quick context on a topic the interviewee mentioned. The Realtime API could pull up relevant data or suggest follow-up questions, helping the journalist keep the interview flowing smoothly. This could enrich live reporting, making it easier for journalists to probe deeper.
As media companies strive to stay relevant and competitive in a rapidly evolving digital landscape, Meta’s celebrity voices and OpenAI’s Realtime API offer significant advantages. These tools go beyond traditional AI applications, providing customizable voices and real-time responsiveness that could enhance audience engagement, streamline reporting, and position newsrooms as leaders in innovative storytelling. These high-level capabilities would allow media companies to tailor their news delivery and interaction with audiences, making news consumption more personalized and immediate.
These advancements are not just about making existing processes faster — they’re about creating new ways for audiences to interact with news. The ability to ask questions in real-time, select a familiar voice to narrate daily briefings, or receive instant updates as stories evolve could make the news feel more immediate, accessible, and relevant.
The future of journalism lies in properly leveraging these AI-driven technologies to meet audience demands for immediacy and customization. As newsrooms explore these cutting-edge capabilities, they have the potential to reshape the way people experience news, making it more relevant, accessible, and engaging.
As the next wave of AI tools arrives, newsrooms and media companies have the opportunity to redefine their relationship with audiences. These new technologies could bring a level of engagement that traditional methods struggle to match. Newsrooms that are willing to experiment with and adopt these technologies will be better positioned to engage audiences.
Michael Crystal is a senior at Northwestern University pursuing a dual major in computer science and journalism. On campus, he works in the Knight Lab and writes for the Generative AI in the Newsroom blog, exploring the intersection of technology and media. His professional experience includes software engineering roles at Pinterest and AtScale, where he honed his skills in developing complex software systems. Feel free to reach out to him at [email protected].