RootByte Just Added Podcast Mode. Two AI Hosts Will Discuss Any Article — While You Listen.
Every article on RootByte now has a play button. Click it, and two AI hosts named Alex and Sam will have a real conversation about the story — complete with reactions, jokes, and genuine surprise. No pre-recording. No human voice actors. Generated in real-time.
Key Takeaways
- •Every article on RootByte now has an AI podcast — click play and two hosts discuss the story in real-time
- •The pipeline: Mistral AI writes the conversation script, Google Gemini voices it with two distinct human-like speakers
- •Generation takes 15-30 seconds — no pre-recording, no human voice actors, no editing
- •Audio is cached for 24 hours after first generation — instant playback for subsequent listeners
Root Connection
From radio dramas in the 1920s to podcasting's 2004 birth to AI-generated conversations in 2026 — the evolution of audio storytelling just took its wildest leap yet.
Podcast Listeners Worldwide (billions)
Source: Statista / Edison Research
Timeline
KDKA Pittsburgh broadcasts one of the first commercial radio programs. Families gather around a wooden box to hear voices from nowhere. The idea that audio can inform and entertain takes root.
Orson Welles broadcasts 'War of the Worlds' so convincingly that listeners panic, thinking aliens have invaded. Audio storytelling proves it can be more immersive than any other medium.
Apple launches the iPod. For the first time, thousands of hours of audio fit in a pocket. The hardware for a podcast revolution is ready — the word 'podcast' hasn't been invented yet.
Former MTV VJ Adam Curry and software developer Dave Winer create iPodder, software that automatically downloads audio to iPods via RSS. Ben Hammersley coins the word 'podcast' in The Guardian.
Serial debuts — a true crime podcast that becomes the fastest podcast to reach 5 million downloads. Podcasting goes mainstream. Everyone from celebrities to your neighbor starts a show.
Google launches NotebookLM's Audio Overview — upload documents and two AI hosts discuss them. The concept of AI-generated podcasts enters public consciousness.
RootByte ships Podcast Mode — every article gets an on-demand AI podcast. Two hosts (Alex and Sam) discuss the story in real-time using Mistral AI for scripting and Google Gemini for voice synthesis.
Open any article on RootByte. Look just below the share buttons. You'll see a player with a purple play button and three words: "Listen to this article."
Click it.
Within about 20 seconds, two voices will start talking. Not reading. Talking. One voice — that's Alex — will lay out the facts of the story with the energy of someone who just discovered something incredible and can't wait to tell you about it. The other voice — Sam — will react the way you'd react: with surprise, follow-up questions, jokes, and the occasional "Wait, are you serious?"
They'll have a conversation about the article you were about to read. Except now you can listen to it while you cook, commute, walk the dog, or pretend to work.
The best podcasts don't feel like someone reading to you. They feel like eavesdropping on two friends who happen to know a lot about something cool.
— RootByte Editorial Team
This is Podcast Mode. And as of today, every single article on RootByte has it.
HOW IT WORKS
The technology behind Podcast Mode is a two-step pipeline that runs entirely in the cloud, on demand, in real-time.
Step one: the article's content gets sent to Mistral AI — one of Europe's most powerful language models. Mistral doesn't just summarize the article. It writes a complete podcast script — a natural conversation between two hosts named Alex and Sam. Alex is the knowledgeable one who presents facts with enthusiasm. Sam is the curious one who asks the questions you'd ask, reacts with genuine surprise, and keeps things fun.
The script includes real reactions. "No way." "That's wild." "Hold on, let me get this straight..." It's designed to feel like two friends talking at a coffee shop, not two robots reading a teleprompter.
Step two: the script gets sent to Google Gemini's multi-speaker text-to-speech engine. Gemini doesn't just convert text to audio — it generates two distinct human-like voices with natural intonation, pacing, and emotion. Alex gets one voice (Kore — warm and authoritative). Sam gets another (Puck — bright and reactive). The result sounds like a real podcast.
We didn't want a robot reading an article. We wanted two hosts who react, interrupt each other, laugh, and say 'Wait, seriously?' — because that's what makes audio compelling.
— RootByte development notes
The entire process — from clicking play to hearing the first words — takes about 15 to 30 seconds. After that first generation, the audio is cached for 24 hours. Anyone else who clicks play on the same article gets instant playback.
No human voice actors were hired. No audio was pre-recorded. No editing was done. It's generated fresh, every time, from the article you're looking at.
WHY WE BUILT IT
Here's the truth about online reading: most people don't finish articles. Studies consistently show that the average reader gets through about 20-30% of a web article before bouncing. That's not because the content is bad — it's because reading on a screen competes with every notification, tab, and distraction in the universe.
Audio changes the equation. When you're listening to a podcast, you can do other things simultaneously. Your hands are free. Your eyes are free. The content travels with you. And there's something about hearing two people get excited about a topic that makes it more engaging than reading the same information on a screen.
The future of media isn't choosing between reading and listening. It's having both, instantly, for every piece of content ever published.
— RootByte Editorial Team
We looked at what Google did with NotebookLM's Audio Overview feature — where you upload documents and two AI hosts discuss them — and thought: why should this require uploading anything? Why can't every article just have this built in?
So we built it in.
WHAT MAKES IT DIFFERENT
NotebookLM is incredible, but it's a research tool. You upload PDFs, paste URLs, and it generates an overview. It's designed for students, researchers, and professionals processing large amounts of information.
RootByte's Podcast Mode is different. It's integrated directly into the reading experience. You don't upload anything. You don't leave the page. You don't need an account. You just click play on any article and listen.
The conversation is also specifically tuned for RootByte's editorial style. Our articles trace technology back to its historical origins — and the AI hosts do the same thing. They don't just summarize. They react to the history, connect the dots, and highlight the most surprising facts. If an article mentions that Bluetooth is named after a Viking king, you can bet Sam is going to lose it over that fact.
THE TECH STACK
For the technically curious, here's exactly what powers Podcast Mode:
Script generation uses Mistral's small model with a custom system prompt that enforces the two-host format, natural conversational patterns, and a strict word limit. The model receives the article's title and content (capped at 3,000 characters to stay within token limits) and returns a dialogue-formatted script.
Voice synthesis uses Google Gemini 2.5 Flash Preview TTS with multi-speaker configuration. Each speaker gets a distinct prebuilt voice. The API returns raw PCM audio at 24kHz, which we convert to WAV on the server and stream directly to the browser.
The frontend player is a custom component with play/pause, a progress bar with seek functionality, and real-time duration tracking. It uses the Web Audio API through a standard HTML5 audio element — no external audio libraries.
Total cost per podcast generation: fractions of a cent. Literally. Mistral's input is a few hundred tokens, Gemini's TTS output is a few thousand. At current API pricing, we could generate tens of thousands of podcasts for the cost of a single cup of coffee.
WHAT'S NEXT
This is version one. We're already thinking about what comes next.
Episode memory — so you can pick up where you left off. Playlist mode — queue up multiple articles as a continuous podcast feed for your commute. Speed controls — 1x, 1.5x, 2x playback. Download as MP3 — take it offline. Custom host personas — maybe you want a more serious tone, or a more comedic one.
And the big one: user-submitted topics. Imagine typing a question into RootByte — "What's the history of encryption?" — and getting a full podcast episode generated on the spot, with research, facts, and two hosts walking you through it.
We're a small, independent newsroom. One human, one AI. No funding. No investors. And we just shipped a feature that turns every article into a podcast with two AI hosts who sound like they actually care about what they're discussing.
Click play. Listen to this article. Let Alex and Sam tell you about themselves.
How did this make you feel?
Recommended Gear
View all →Disclosure: Some links on this page may be affiliate links. If you make a purchase through these links, we may earn a small commission at no extra cost to you. We only recommend products we genuinely believe in.
Framework Laptop 16
The modular, repairable laptop that lets you upgrade every component. The right-to-repair movement in action.
Flipper Zero
Multi-tool for pentesters and hardware hackers. RFID, NFC, infrared, GPIO — all in your pocket.
The Innovators by Walter Isaacson
The untold story of the people who created the computer, internet, and digital revolution. Essential tech history.
reMarkable 2 Paper Tablet
E-ink tablet that feels like writing on real paper. No distractions, no notifications — just thinking.
Keep Reading
Want to dig deeper? Trace any technology back to its origins.
Start Research