MumbleFlowGet MumbleFlow — $5

Aura Technologies · native Mac software

Words should feel finished.

MumbleFlow is a voice workspace for the gap between what people say and what they actually need: writing that is ready to send, and meeting knowledge that is easy to trust, find, and act on.

Apple Silicon · macOS 26+ · English · Models download during setup

One app · two workflows

MumbleFlow writes. Oat remembers.

MumbleFlow dictation

From speech to polished writing

Hold or lock Fn, speak with pauses and corrections, and receive final text with fillers, restarts, punctuation, paragraphs, and clearly introduced lists handled on device.

Explore dictation →

Oat meeting notes

From conversation to cited action

Record by choice, follow a live speaker-separated transcript, catch up on the last five minutes, and leave with decisions, next steps, open questions, and links to the supporting moments.

Explore Oat →

Why it is built this way

Fast, grounded, and under your control

01

Natural in

The interface should let people ramble, pause, correct themselves, and think aloud instead of learning dictation syntax.

02

Finished out

Cleanup should improve structure without silently changing names, numbers, links, substance, or intent.

03

Evidence attached

A generated meeting claim is useful only when a person can reach the source that supports it.

04

Recording by choice

There is no ambient buffer, automatic calendar start, or meeting bot. Capture begins after an explicit action.

05

Local intelligence

Customer speech, writing, summaries, embeddings, and meeting search stay on the Mac instead of feeding a hosted inference service.

06

Visible limits

Unavailable connectors, model requirements, retention rules, and uncertain speaker identity should be stated rather than hidden behind marketing copy.

Implementation

A native Apple Silicon application

MumbleFlow is written in Swift 6. SwiftUI powers the full workspace, while AppKit handles the global event layer, accessibility insertion, focus, floating panels, and Mac-specific capture behavior. ScreenCaptureKit records meeting audio without registering a video output.

FluidAudio supplies the on-device speech stack: Silero voice activity detection, English Parakeet TDT v2 recognition, and remote-speaker analysis. Apple's SystemLanguageModel is the preferred semantic engine; an optional Qwen3 1.7B model runs locally through MLX when Apple Intelligence is unavailable.

Current release

Built for a focused first platform

Product information covers release 0.1.4, documented September 3, 2026.

The current download supports Apple Silicon Macs running macOS 26 or later and recognizes English speech. It is distributed as a Developer ID-signed, Apple-notarized DMG. Intel Macs, Windows, Linux, mobile platforms, multilingual parity, cloud sync, and a hosted developer API are not part of this release.

External Gmail, Google Calendar, and Slack source connections are also not enabled in the current download. For setup help or a release-specific question, visit Support or contact the team.

Questions

Frequently asked questions

Who makes MumbleFlow?

MumbleFlow is built by Aura Technologies LLC. Product questions and support requests can be sent to contact@aura-technologies.co.

What is Oat?

Oat is the meeting workspace inside MumbleFlow. It turns explicit meeting recordings into a timestamped transcript, speaker-aware catch-up, cited summaries, decisions, next steps, open questions, and searchable local knowledge.

Is MumbleFlow a browser extension or meeting bot?

No. It is a native Apple Silicon Mac application. Dictation works at the system level, and meetings are captured from selected computer audio and the microphone only after the user explicitly starts recording. No bot joins the call.

What technology does MumbleFlow use?

The app is written in Swift 6 with SwiftUI and AppKit. Its local speech stack uses FluidAudio, Silero VAD, and Parakeet TDT v2. Semantic writing and notes prefer Apple Foundation Models with an optional local Qwen3 fallback through MLX.

Keep exploring

The rest of MumbleFlow

Looking for technical detail? Read the local speech-to-text guide.