RunReplay Studio- Turn Your Apple Health Runs into Narrated Tour-Guide Videos

Posted on May 31, 2026 at 11:10 PM

RunReplay Studio: Turn Your Apple Health Runs into Narrated Tour-Guide Videos

You finished a run along the river, or wandered a new city on foot, and your watch quietly recorded every step. That GPX track is a story — but it usually dies in an app as a thin line on a map.

RunReplay Studio turns that line into a story worth watching.

Drop in a GPX file (or your whole Apple Health export), and the app builds a publish-ready replay video: an animated avatar runs your exact route over a live map, and as it passes real landmarks, a neural voice narrates their history — like a tour guide walking the route beside you. The finished clip exports as a frame-accurate MP4, entirely in your browser.


It’s a tour guide, not just a route animation

This was the whole idea. Plenty of tools can draw a dot moving along a track. RunReplay goes further: it figures out which famous places you actually passed and tells you about them.

Here’s how it works under the hood:

  1. Parse your route. GPX from Apple Fitness, Strava, Garmin — anything standard. Missing timestamps or elevation? It fills in sensibly.
  2. Find the landmarks. It queries OpenStreetMap for notable places near your path — museums, monuments, temples, towers, parks — preferring the famous, Wikipedia-tagged ones.
  3. Tell their story. For each landmark it pulls a real description and a photo straight from Wikipedia, then writes a tour-guide narration line: “Coming up, Tan Si Chong Su. Ancestral Hall of the Tan Clan, a Chinese temple built between 1876 and 1878…”
  4. Bring it to life. The narration is spoken by a neural voice, the landmark photo slides onto the screen, and the avatar talks as it passes.

If a stretch of your route is quiet and passes nothing famous, it degrades gracefully — you still get clean checkpoints, just without the stories.

What’s in the box

  • 🗺️ Live map replay — your avatar runs the route on a clean, colorful basemap, with landmark pins you can tap for a photo and a story.
  • 🎙️ Real narration — historical blurbs and photos sourced live from Wikipedia & Wikimedia, spoken by a neural voice. Pick your guide’s voice (US / UK, male / female).
  • 🎬 Cinematic intro — the route draws itself on, a title card appears, then a smooth follow-cam takes over for the run.
  • 🏷️ Smart titles — the app reverse-geocodes your route into a friendly name like “Tiong Bahru Run” (and you can rename it before exporting).
  • 🍎 Apple Health friendly — upload your export.zip directly; if it contains many workout routes, just pick the one you want.
  • 🎵 Make it yours — choose an avatar, drop in a music bed, or replace the map with your own background image.
  • 📲 Export anywhere — Shorts/Reels (9:16) or YouTube (16:9), 8–30 seconds, as a frame-accurate H.264 + AAC MP4.

Your data never leaves your machine

This is the part I’m proudest of. RunReplay Studio is 100% client-side:

  • No server. Everything — parsing, rendering, voice synthesis, video encoding — runs in your browser tab.
  • No API keys, no accounts. It uses only free, open services: OpenStreetMap Overpass for landmarks, Wikipedia/Wikimedia for stories and photos, and Nominatim for place names.
  • No uploads. Your GPS track isn’t sent to anyone. Your run stays yours.

The MP4 is encoded right on the page using the browser’s WebCodecs API — the same low-level video tech that powers modern web video — so exports are frame-accurate instead of screen-recorded.

A few honest notes

  • Use a Chromium browser (Chrome/Edge) for MP4 export. WebCodecs isn’t everywhere yet; other browsers fall back to WebM.
  • First export downloads the neural voice model (~once, cached after). If it can’t load, the video still exports with on-screen captions.
  • Landmark richness depends on your route. A run past a heritage district lights up with stories; a quiet suburban loop will have fewer.

Under the hood (for the curious)

It’s a no-bundler, no-build app — just ES modules and a handful of CDN libraries. The logic core (GPX parsing, distance math, landmark placement, narration text) is pure and unit-tested; the browser-only bits (Leaflet map, neural TTS, canvas compositing, WebCodecs muxing) sit on top. Motion is distance-based, so the avatar moves evenly even when GPS samples are uneven, and every frame is rendered then encoded one-by-one for a deterministic result.


Take your next run for a spin

Your runs already tell a story. RunReplay Studio just gives them a voice.

Load the demo loop to see it in action, then drop in your own GPX and watch your route become a narrated tour.