Why RoadWhisper Has No Pre-Recorded Stories — and How It Narrates Places On the Spot
Every other audio guide app ships with a fixed library of scripts someone recorded in advance. RoadWhisper ships with none. Here's what it does instead, why that trade-off is deliberate, and what actually happens between the moment you pass a landmark and the moment you hear a story about it.
The problem with a pre-recorded library
A conventional audio guide is, at its core, a media player attached to a list. Someone — a museum curator, a tour company, a city tourism board — decided in advance which stops matter, wrote a script for each one, recorded it, and shipped that fixed bundle inside the app. It works well for the handful of famous sites that justified the effort: a cathedral, a well-known battlefield, the ten stops on a walking tour.
It breaks down the moment you're somewhere that bundle didn't anticipate. Drive through a region that wasn't on the original list — which, on any real road trip, is most of the drive — and the guide simply has nothing to say. It's not that the content is bad; it's that it's finite, and the finite list was fixed the day someone stopped recording.
RoadWhisper's alternative: nothing is pre-written
RoadWhisper doesn't have a list of stops, a script library, or a "coverage area." It has three things running continuously while you drive: a location feed, two public data sources, and a language model that turns whatever those sources return into a short spoken story — composed fresh, every single time, for wherever you actually are.
What's near you, right now
RoadWhisper checks OpenStreetMap for points of interest around your live position — castles, churches, ruins, viewpoints, monuments. The search radius scales with your speed, so it looks further ahead on a motorway than on a slow village road, and nothing is looked up until you're actually near it.
What's publicly known about it
For each place found, RoadWhisper pulls whatever public facts exist — mainly from Wikipedia. If a place has no available facts anywhere, RoadWhisper skips it rather than asking the AI to invent a history for it out of nothing.
Turning facts into a spoken story
A language model writes a short narration grounded in those facts, in your chosen tone and language. A second pass then rewrites any numbers — years, centuries, counts — as naturally spoken words instead of digits, so "1628" is actually said aloud the way a person would say it, not read out as four separate digits.
Why the two-step writing process exists
The number-spelling step sounds like a minor detail, but it's the difference between narration that sounds natural and one that sounds like a screen reader. Text-to-speech engines are inconsistent at reading digits aloud in a grammatically correct way — especially in languages with case systems, where "in 1628" needs a different word ending than "the 1628 fire" would. Asking a single AI pass to both narrate the history and get every number's grammar right in the target language turned out to produce mistakes in one or the other. Splitting it into two focused passes — one that knows the facts, one that only fixes number formatting — fixed both.
Because every story is generated at the moment you pass a place, RoadWhisper needs an internet connection to work — it has no offline script bundle to fall back on. Places that get passed often are cached after the first generation, so a popular stretch of road narrates faster on the second trip through.
What this trade-off actually buys you
- No blank stretches. A minor roadside chapel or a small-town war memorial will never justify a professional recording budget — but if it's on the map and has a public source of facts, RoadWhisper can still say something about it.
- Nothing to plan. There's no route to select or stop list to browse beforehand — the app reacts to wherever you happen to be driving, on any road, in real time.
- One story, many tones. The same place can be narrated as a straightforward historical account, a humorous aside, or a simplified version for kids — because nothing was locked into a single fixed recording in the first place.
- The honest limitation. A generated narration is only as good as the public facts behind it, and it's still AI-written text — treat specific dates or figures as a starting point for curiosity, not a verified historical record.
Hear it for yourself
Free to download, no account needed. Press Start and drive — RoadWhisper narrates whatever's actually along your route.
Frequently asked questions
So there's really no fixed content anywhere in the app?
Correct — every story is composed at the moment you're near a place, grounded in OpenStreetMap and Wikipedia data rather than pulled from a pre-written script.
What happens if a place has no Wikipedia article?
RoadWhisper skips it rather than generating an invented history — narration only happens for places where a real public source of facts is available.
Why does it need an internet connection?
Because stories are generated live, not played from a downloaded file. There's no offline bundle — the location lookup, the fact-finding, and the AI writing all happen in real time as you drive.
Does this mean stories are different every time?
The exact wording can vary between generations, though frequently passed places are cached after the first generation so a familiar stretch of road narrates consistently and faster on repeat drives.