Tools

Building the transcription app keeps producing small, useful things that are not the app. This is where they live: one job each, running entirely in your browser on your own key — and each one carries a JS API, so an agent can use it exactly as a person does.

Available now
VIDEO

🎬 Extract audio from a video

Drop in an .mp4, .mov or .webm and take the soundtrack out as an .m4a — then send it straight to the app to be transcribed, without a trip through your downloads folder.

Free, and nothing is uploaded: FFmpeg runs as WebAssembly in your own tab, so there is no key, no model and no cost. The first run fetches 32 MB of it, once.

JS API · manifest.json

AUDIO

🎙 Text to speech

Paste any text, pick one of six voices, and download the spoken .wav. Built for reading the Journalist's updates aloud as news-style voice memos, but it reads whatever you give it.

One HTTPS call to OpenRouter with your key; the cost is read back from the generation, not estimated.

JS API · manifest.json

MEDIA

🎙 Podcast & YouTube transcriber

Paste a podcast episode or a YouTube link and get the transcript back, plus a summary, key points and topics.

Not browser-only, on purpose: a podcast feed and a YouTube transcript can't be fetched from this origin (CORS), so this one runs through a transcription service instead of your own key. No cost to you either way — see the note on the page for exactly what that means.

manifest.json

Already elsewhere on this site
AUDIO

The transcription app

The product itself: a WhatsApp voice note in, a transcript, an analysis and a short brief out. It is the reason everything here exists, and it publishes the same kind of JS API these tools do.

JS API · manifest.json

DESIGN

Design candidates

Three interface themes and four culture packs, each running the real flow, side by side. Not a tool you would use daily — a way to see and argue about what the app should look like.

What makes something a tool here

  • One job, one page. If it needs a tour, it is a product, not a tool.
  • No backend, with one named exception. Everything here runs in the browser — except the podcast & YouTube transcriber, which cannot: neither source sends CORS headers for this origin, so it runs through a transcription service instead. It is the only tool that does, and it says so on its own page.
  • Your key, your bill. Anything that costs money uses your own OpenRouter key, kept in this browser and sent only to OpenRouter.
  • An API, not just a UI. Every tool publishes window.__tool with a manifest and a skills document, so a Playwright script or an agent can do what the page does without a human clicking.
  • Honest about cost and limits. Where a real number exists, we read it back rather than estimate it.