Automate your dictation: run Shortcuts, webhooks, and files when you finish speaking

Dictation usually ends the same way: the text lands in whatever app you’re in. But often that’s just step one — you wanted the note in your journal, the task in Things, the idea POSTed to a webhook that kicks off a workflow. OpenWhisp can now do that automatically, the moment you stop speaking.

Output targets: send a dictation somewhere other than the focused app

In Settings › Output › Output target you pick where the final transcript goes:

• A file — append each dictation to an Obsidian daily note, a Logseq journal, or any .md/.txt, with an optional dated heading.
• A macOS Shortcut — hand the text to any Shortcut you’ve built (add to Reminders, run AppleScript, whatever the Shortcuts ecosystem can do).
• A webhook — POST the transcript as JSON (text, language, app, timestamp) to Notion, Zapier, n8n, or your own endpoint, with optional auth headers.

Rules: trigger actions based on what you said

Rules (Settings › Rules) go a step further: they run actions when a transcript matches. A rule is a match condition — starts-with, contains, exact, or regex, optionally scoped to one app — plus an ordered list of actions: insert a snippet, open a URL (with {{text}} for the transcript), run a script, run a Shortcut, POST a webhook, or append to a file. Say the right trigger phrase and the automation fires.

Drive OpenWhisp from a launcher: the openwhisp:// scheme

OpenWhisp answers an openwhisp:// URL, so any launcher or script can control it — no CLI needed:

open "openwhisp://record"                       # start / stop dictation
open "openwhisp://paste-last-result"            # paste the last result
open "openwhisp://refine?instruction=make%20it%20formal"

# chain verbs — they run in order, all-or-nothing
open "openwhisp://?switch-mode=email&record"

Point Raycast or Alfred at one of these and bind it to a hotkey. The scheme deliberately exposes only a small, validated allow-list of safe verbs — never a shell string or an arbitrary file path — and rejects anything malformed as a whole.

Nothing gets dropped

The important design choice: every one of these is a fail-open side channel. If a webhook is down, a Shortcut errors, or a file can’t be written, your words still land in the focused app exactly as they always did. An automation can never cost you a dictation. And it’s all local by default — Rules run on dictation only, so an agent’s transcript never triggers them unless you opt in.

For the tinkerers

This is OpenWhisp leaning into its hackable, open-source nature: your voice becomes a trigger for whatever you can wire up on your own machine, with the safety of knowing the plumbing can only ever add to normal dictation, never break it.