Get from install to your first words.

MumbleFlow is a $5 one-time purchase with no recurring app subscription. Checkout uses Stripe, and the completed payment opens a verified download for the Developer ID-signed, Apple-notarized DMG.

A paid download link can be redeemed once. If it has already been used, your payment cannot be verified, or the download fails, email contact@aura-technologies.co for help or another download.

1. 01

   Install the app

   Drag MumbleFlow from the DMG into Applications and open that installed copy. The main window opens at launch; after closing it, choose Open MumbleFlow from the menu bar to reopen it.

2. 02

   Allow the microphone

   Microphone permission is required for dictation and for your side of a meeting. Confirm the input meter moves when you speak.

3. 03

   Allow Accessibility

   This lets the global Fn gesture work across apps and lets MumbleFlow insert the finished result into the focused editor.

4. 04

   Download speech models

   Complete the verified English speech-model download while online. Everyday speech recognition can then run without a network connection.

5. 05

   Test in a simple editor

   Open Notes or TextEdit, focus a document, hold Fn, say a short sentence, and release. Confirm the result is inserted and available on the clipboard.

• Quick thought: hold Fn, speak, and release to process immediately.
• Longer dictation: double-tap Fn to lock recording, then press Fn once when you are finished.
• Cancel: press Escape while recording to discard the capture without inserting.
• Shortcut conflict: choose Option-Space, Control-Space, or Shift-Command-Space as a fallback in Settings.

The frosted HUD shows a live waveform only after audio capture starts. A spinner replaces it during transcription and polishing, with a small Apple Intelligence or Qwen badge showing the active semantic engine. Quiet sound cues mark start, lock, and finish.

MumbleFlow targets the field focused at insertion time. Native editors prefer Accessibility selected-text replacement. Browser and custom editors use a focused Command-V event so modern web surfaces—including canvas-style document editors—can receive normal paste input.

If a destination still rejects insertion, the final result stays on the clipboard and can remain in encrypted Dictation History. Confirm Accessibility permission belongs to the installed copy, not an older development build, then restart the destination app and MumbleFlow.

1. 01

   Press Record Meeting

   MumbleFlow never starts from the calendar or an ambient buffer. Choose all computer audio or a specific running app.

2. 02

   Grant Screen & System Audio

   Open System Settings → Privacy & Security → Screen & System Audio Recording and enable the installed MumbleFlow app.

3. 03

   Restart MumbleFlow

   macOS may require a full app restart before a newly granted capture permission becomes effective.

4. 04

   Check both channels

   Your microphone should appear as You; selected computer audio is the remote channel. No video stream is captured or retained.

For the complete capture and speaker workflow, read meeting transcription with speaker labels.

Parakeet performs English speech recognition. Apple Intelligence or optional Qwen handles conservative long-form rewriting and generated meeting intelligence. Dictations shorter than ten lexical words skip the semantic model to reduce delay while still receiving deterministic cleanup.

If Apple Intelligence is unavailable, Settings explains why and offers the approximately 930 MB Qwen3 fallback. Personal-dictionary terms of at least three characters can also use an optional approximately 98 MB acoustic rescoring model. Model downloads are checksum-verified before use.

Enable “Keep the latest test recording” in Settings only when you need to diagnose recognition or polishing. MumbleFlow retains one microphone WAV plus a report containing raw transcript, final text, and processing mode. A new test replaces the old one; turning diagnostics off deletes both files.

Still stuck? Copy the comparison without confidential content and email contact@aura-technologies.co. Include your MumbleFlow version, Mac model, macOS version, and the destination app.