DEContact
Flowwy

Offline dictation for macOS

Hold fn, speak, let go.

Flowwy is voice input for macOS that sends nothing anywhere. Speech recognition and the optional AI cleanup run on your Mac. Hold fn, speak, let go, and the text is where your cursor is, in any app. Free, no account, no telemetry.

Free, no account

The Flowwy home screen: word count, words per minute and day streak, with today's dictations below showing app, time, latency and model.

What it's about

Dictating is faster than typing, but most dictation apps send your voice to the cloud. For emails to clients, notes from conversations or code comments, that is not a good idea.

Flowwy recognizes speech with whisper.cpp directly on your Mac's GPU (Metal), with models from Whisper Tiny to Large v3 Turbo and 99 languages with automatic detection. From releasing the key to text takes about 0.4 seconds (measured on an M1 Max with a small model). If you want, a local language model then tidies up the text: punctuation, filler words, self-corrections.

Privacy you can check for yourself: under Settings › Privacy, every network request is listed with host, purpose and bytes. Audio stays in memory, models are only downloaded when you click. There is no account and no telemetry.

How it works

  1. Hold fn

    Or right ⌥, ⌘ or ⌃. In any app, wherever your cursor is.

  2. Speak

    Whisper recognizes the speech locally on your Mac.

  3. Let go

    The text appears at the cursor position, cleaned up if you want.

The Transforms view in Flowwy: select text, press ⌥1, review the changes, plus cards for Polish, Prompt Engineer and custom transforms.

What Flowwy can do

Works in any app with a text cursor, with special handling for editors, terminals and coding agents. Optional AI cleanup via local models.

On the device

Speech recognition with whisper.cpp on Metal. Models from 78 MB to 1.6 GB, smaller ones are faster.

Your own models

Cleanup via Ollama, llama.cpp, LM Studio or MLX. For speech, any local OpenAI-compatible server works.

Context-aware

Reads the app name and the text around the cursor for spacing and spelling. Can be turned off.

Dictionary

Your own names and technical terms, plus correction rules like "kuber netties" to Kubernetes.

Snippets and styles

Say a trigger word and get the whole text. Styles from formal to raw.

Transforms

Select text, press a shortcut, and the result replaces the selection: Polish, Prompt Engineer or your own instructions.

For coding agents

Recognizes Claude Code, Codex, Gemini CLI, Cursor and more, plus editors and terminals. "slash compact" becomes /compact.

Scratchpad

Voice notes that stay on your Mac.

Privacy log

Settings show every request the app makes.

Privacy and security

Speech
entirely on your Mac
Account
none, no telemetry
Network
every request in the log
Endpoints
local only, remote blocked unless you allow it
Flowwy Cloud
optional, off by default, text only
System
macOS 13 or newer, Apple Silicon (M1 or newer)

Pricing

As of October 2026. Current prices at flowwy.dev. The app is free, and so is Flowwy Cloud during the beta.

Flowwy

Free

no account

  • Dictation, dictionary, snippets, styles
  • Local models from 78 MB to 1.6 GB

Flowwy Cloud (Beta)

Free

during the beta

  • Optional, off by default
  • Cleanup and transforms on hosted models, speech stays local

Who Flowwy is for

  • Heavy writers on the Mac

    Emails, notes and documents faster than typing them.

  • Developers

    Prompts, commit messages and comments by voice, with snippets and detection for coding agents, editors and terminals.

  • Professions bound by confidentiality

    Dictate without anything leaving the computer.

Frequently asked questions

Does Flowwy send my voice to the internet?

No. Speech recognition and cleanup run on your Mac. The only things that go out are a model download when you start one, and an update check you can turn off, with no audio, text or identifiers.

Does Flowwy work offline?

Yes, as soon as a speech model is loaded. The cleanup is optional.

What does Flowwy cost?

Nothing. The app is free and needs no account. Flowwy Cloud is also free during the beta.

Which permissions does Flowwy need?

Microphone to listen, Accessibility to type into other apps, and Input Monitoring for the fn key.

Do I need an AI model?

Not for speech recognition, that is built in. For the optional cleanup you can connect a local model via Ollama, llama.cpp, LM Studio or MLX.

Which languages does Flowwy support?

The multilingual Whisper models detect 99 languages automatically. Whisper Base (English) is a smaller option for English only.

What is Flowwy Cloud?

An optional beta, off by default. It runs cleanup and transforms on hosted models, while speech stays local. Only text is sent. The gateway runs with us in the EU, but during the beta requests go to model providers that may be located outside the EU. For sensitive content, use the local cleanup.

Does Flowwy run on Intel Macs?

No. Flowwy needs macOS 13 or newer and Apple Silicon (M1 or newer).

Questions about Flowwy?

Plans, privacy, integration or feedback: write to us, and we'll reply personally.