ABOUT THE PROJECT
Local News Matters is testing an innovative way to bring you the daily news, so take a moment to listen in!
We are launching a brand-new podcast — available every afternoon — using our original, independent reporting combined with an AI experiment to mimic the voice of Bay City News editor Leslie Katz. Tune in each day for quick updates on the biggest stories and events happening around the Bay Area!
This experiment is all about learning and gaining feedback from the community about what is useful to our readers and listeners. That’s where you come in! Each podcast will feature a survey. Let’s hear what you think!
The Daily News Roundups used in this project are created once every day, based on news articles created by human reporters and editors at Bay City News. For this project, we prompted ChatGPT to analyze the articles produced by our staff during this 24 hour period and to choose 5 stories to highlight based on newsworthiness and human interest, according to the AI tool. We prompted ChatGPT to summarize these 5 stories into a script suited for podcast narration. Then we used ElevenLabs and other tools to help us convert the text into audio based on the voice of Leslie Katz, one of our Bay City News editors. This content was verified by a human editor.
Latest Episodes
Check here every day for roundups of the key local news that matters to the greater Bay Area, provided in both audio and written formats!
ABOUT THE PROCESS
Part 1: Experimenting with AI tools to turn written content into daily podcasts
Step 1: The content
LocalNewsMatters.org is a public-service news site operated by the nonprofit Bay City News Foundation (BCNF), which is also affiliated with the Bay City News Wire Service. This wire service provides news coverage to media outlets across 13 counties, from Mendocino to Monterey. Each day, our editors compile a news roundup for our media clients. Through this project, we’re making that same roundup—minus the paywall-protected full stories—available to our Local News Matters readers.
Step 2: The technology
We are using readily available AI voice tools like ElevenLabs, Speechify and Everlit to test the best way to convert our top stories into an AI-narrated podcast. Other tools we might try include Play.ht, Descript, or a bespoke tool that can use a real staff voice to read news roundups that change each day.
Step 3: The product
Our aim is to create a daily podcast that provides an alternative news platform for our existing readers and an effective product to attract a younger 18-25 audience that is more drawn to audio than text. Using AI will help us turn our existing original journalism into alternative forms to build audience, preserve our staff time to do more original journalism, and potentially create a new way to monetize our work via sponsorships or membership donations.
The result? A long, boring podcast.
Our first attempt to turn the daily news roundups into audio content resulted in a product that was too long, too repetitive, and too boring. We learned that the requirements for audio were different from the requirements for written content. While this might appear to some as a failed first attempt, we are excited about the opportunity this gives us to improve our content, understand the tools better, and iterate to create a better product. To solve this we are going back to the drawing board and considering different strategies to transform our content more successfully.
Here is an example:
Part 2: Refining the content to create a product worth listening to
Step 1: Translating the content
To refine the process, we explored three different approaches using our original reporting from March 27, 2025. By trying each of these approaches on the same day’s source material, we can compare both the results and the workflow of each approach. Examples are linked below:
A) Having a human editor tailor the content for audio (is this “failure” a chance for our team to refine our content for broadcast clients?)
B) Using ChatGPT to take the original news summary and make it more suited for a news narration, not just for reading (can we ask AI to turn our existing roundups into podcast scripts?)
C) Leveraging AI to highlight 5 stories from our day’s coverage, surfacing articles based on newsworthiness and human interest (what happens if we turn the process of summarizing the news over to AI?)
Each method offers different levels of control and efficiency, and testing them will help us determine the best way to deliver a concise, engaging audio product. Our next step is to get feedback from both our audience and our own staff to determine which version listeners prefer, as well as which version works best in terms of efficient workflow. Our goal is to create a sustainable daily audio product that serves our audience.
Step 2: Trying again and improving our workflow
With the content improved, we will then repeat the steps of part 1: turning that content into audio and building a daily workflow to offer our news roundups as audio. As we did in our first attempt, the engineering team will combine the human-voiced AI disclaimer from our staff editor Leslie Katz with the News Roundup text-to-audio conversion on the ElevenLabs platform, using background recordings provided by Leslie to create the new audio, a verbatim “reading” of the Roundup text. This time, we are going to repeat that process using an improved version of the podcast, based on the response we get to the three variations we tried.
Step 3: Imagining the potential
Once the process of producing these daily roundups has been refined and improved, we can begin to imagine all the potential of what we can do next! Here are a few possibilities we are exploring:
1) Automating the delivery of this content to our readers. Whether as daily newsletters or audio uploads pushed directly to subscriber devices, we can bring these daily news round-ups right to readers and listeners.
2) Exploring how else we can utilize this audio technology with our other content. The options are endless and we want to hear from readers what kinds of content they want to hear in audio formats!
3) Continuing to improve and build upon this work. We hope that by the end of this experiment we will end up with a good daily product, but with the technology developing so quickly we will likely be able to improve upon it even further.