MinardPeople
Cards, exactly when they matter
The first time the narrator names someone, a card steps forward with a portrait and a line of context. One at a time, queued so nothing talks over anything. Later mentions become small chips on the map.
The interactive way to hear history
Listen to history.
Watch it move.
Minard is a desktop app for people who love history podcasts. Play an episode and a map animates alongside the narration, so when the host says “the tanks reached Minsk”, you watch them get there.
- 1 Add an episode
- 2 Minard turns it into a map
- 3 Press play
Minard
Barbarossa: The First Twelve Days
Compiling the atlas…
↓ Scroll to play the episode
0:00 / 7:59
How it works
You add an episode. Minard does the homework.
Before you press play, an AI reads the transcript and plans the map. Once that’s done, playback is instant, and pausing, skipping and 2× all just work.
- 01
Add an episode
Paste a YouTube link or a podcast feed, or drop in a transcript (and the audio, if you have it). Turning an episode into a map uses an AI, so you add your own API key from OpenRouter or OpenAI once. It stays in your computer’s keychain. It’s quick: about a minute and a half for an hour-long podcast, and playback can start before it’s finished.
- 02
An AI plans the map
It reads the narration and decides what to show: who’s involved, where the armies move, when a person deserves a card. It only ever names places. Minard’s own code works out exactly where they are and when to draw them, so the map stays tied to real geography.
- 03
Press play. Or jump anywhere.
Skip ahead ten minutes, drag the timeline, change the speed. The map always matches what’s being said at that moment, because it’s redrawn from the timestamp alone. Try it: drag the slider.
stateAt(342.0)0units0arrows0fronts0pockets0cast
What it does
A documentary camera, a cartographer and a very quiet research assistant.
Honesty
What was said, and what was guessed
Everything the narrator said is drawn in solid ink. Anything Minard had to infer is dashed and fainter. Hide it whenever you like.
Mini player
A map that follows you
Shrink the window to a small always-on-top atlas that sits on every desktop and snaps to a screen corner. Drag this one around.
Keyboard first
Never reach for the mouse
Try them. They light up as you press.
Inspector
Yours to correct
If the AI gets something wrong, you can see every element on the map and switch it off or fix it. Redoing the map never repeats the AI work you’ve already paid for.
- g083map.advance2nd Panzer Group → Minsk
- g086map.encircleMinsk pocket
- g017map.battleOpening barrage · Brest
- g035core.cardHeinz Guderian
- g078map.advanceA “what if” thrusthypothetical
Local & private
A folder, not an account
Episodes are plain files. Sync the folder with Dropbox or iCloud, or send one episode to a friend as a single small file. Your AI key stays in your computer’s keychain.
- Documentary camera that glides and drifts
- Flashbacks tinted sepia
- Speeds from 0.5× to 2×
- Chapters on the timeline
- Split-flap dates
- Wikipedia summaries
- Works with YouTube or your own audio
- Per-episode budget cap
Every era, its own atlas
Same board. Re-inked for the century.
The AI’s plan only says “attacker” and “defender”. The era decides what that looks like, from hatched staff-map arrows to white line on blueprint. This is one moment from the demo episode, drawn five ways.
Staff map · Second World War
Hatched arrows, sea-grey water, a printed staff officer's sheet.
Why “Minard”
Named for the best statistical graphic ever drawn.
In 1869 the French engineer Charles Joseph Minard drew Napoleon’s march on Moscow as a single ribbon whose width was the size of the army. Edward Tufte later called it “may well be the best statistical graphic ever drawn.”
He let one line tell a whole story. Minard, the app, tries to do the same with motion: to let the map do the telling while you listen.
422,000men still marching, as Minard drew it
- ≈ 50¢to map an hour-long podcast
The AI usage for a full episode, paid to your AI provider. You bring the key.
- ≈ 90 secto get it ready
And playback can start even sooner, while the rest finishes.
- 0.5–1.3 MBper hour, shared as one file
A .minard file carries the whole timeline, cards and portraits.
- 24 MBof atlas, all offline
Coastlines, rivers, relief and a place-name gazetteer ship inside the app.
Download
Put a map behind the next episode.
Minard v0.1.0 for your desktop. It plays the built-in Barbarossa episode straight away. To turn your own podcasts into maps, you’ll add an AI key (see below).
- Download
macOS
Apple Silicon · .dmg
v0.1.0 · 163 MB
- Download
macOS
Intel · .dmg
v0.1.0 · 169 MB
- Download
Windows
x64 & ARM64 · installer
v0.1.0 · 258 MB
Linux
AppImage
Not released yet
What you’ll need
- A Mac or Windows PC. That’s it for the built-in demo episode, which needs no setup or key.
- An AI API key from OpenRouter or OpenAI, to turn your own episodes into maps. You pay the AI provider directly for the usage: roughly 50¢ for an hour-long podcast, with a spending cap you set per episode.
- Your key stays with you. It’s kept in your computer’s keychain and only sent to the AI provider you chose.
macOS builds are signed and notarized by Apple.
Windows builds are unsigned for now, so SmartScreen may ask you to confirm before running.
Questions
Things people ask.
Do I need an AI account or key?
To turn your own episodes into maps, yes. An AI reads each transcript and plans the map, so you paste an API key from OpenRouter or OpenAI into Settings. You pay that provider directly for the usage, your key is kept in your computer’s keychain, and each episode has a spending cap so a long one can’t surprise you. The built-in Barbarossa episode plays with no key at all.
What can I use it with?
A YouTube link (Minard fetches the captions), a podcast feed for episodes that publish a transcript, or any transcript file (SRT, VTT and similar), optionally with your own audio file.
How long and how much?
An hour-long podcast comes to roughly 50¢ in AI usage and takes about a minute and a half to get ready. Playback can start even earlier, while the rest finishes, and simply waits if it ever catches up. Your exact cost depends on the AI model you choose.
Does it work offline?
Yes, for anything already compiled. The basemap, gazetteer, card text and portraits are all stored with the episode. YouTube episodes still need a connection for the video itself.
How accurate is the map?
Minard is a companion, not a source. An AI plans the map, so geography can be coarse, especially for 1941 front lines. Anything the narrator didn’t actually say is drawn dashed, and you can switch off or fix any element.
Can I share an episode?
Yes. “Share…” writes one .minard file, roughly 0.5–1.3 MB for an hour. Drop it on the window to import. YouTube episodes play straight away; audio episodes ask the recipient to point at their own copy of the audio.
What about YouTube’s terms?
Captions are fetched with yt-dlp for personal use, which may conflict with YouTube’s Terms of Service. The video plays in YouTube’s own embedded player, which Minard never hides or shrinks below the size YouTube requires.
Which platforms?
macOS (Apple Silicon and Intel, signed and notarized) and Windows (x64 and ARM64, currently unsigned). A Linux build is possible but hasn’t been released.