FreeFlowDownload
Press

Press kit

By Yiftach FreemanUpdated

FreeFlow Dictation is an on-device voice dictation app for macOS: hold ⌥, talk, let go, and the words land in any app, transcribed on the Mac by Parakeet, Whisper or Cohere Transcribe models. $99 once for three Macs, 14-day trial, no account, no telemetry. It also answers Claude Code’s plan, question and permission cards by voice.

Boilerplate

One line. FreeFlow Dictation is on-device voice typing for the Mac: hold a key, talk, and the words appear in any app, with nothing sent anywhere. $99 once.

Short. FreeFlow Dictation turns speech into text in any Mac app, entirely on the Mac. Hold ⌥ and talk; let go and the words land at the cursor. Six speech models run on Apple Silicon’s Neural Engine, so it works with Wi-Fi off and sends no audio to any server. Modes, a personal dictionary and on-device formatting with Apple Intelligence do the tidying. It is $99 once for three Macs, with a 14-day trial and no account. It is also the only dictation app that answers Claude Code’s plan, question and permission prompts by voice.

Fact sheet

NameFreeFlow Dictation (one word, capital F and D). “FreeFlow” alone in prose after the first mention is fine. Not to be confused with FreeFlow by Zach Latta, an unrelated open-source app.
What it isOn-device voice dictation for macOS. Hold ⌥, talk, let go; the words land at the cursor in any app.
Price$99 once, up to three Macs, every update included. 14-day full trial, no card, no account.
RequirementsmacOS 14 or later on Apple Silicon. Cohere Transcribe needs macOS 15; Apple Intelligence formatting needs macOS 26.
Speech modelsNVIDIA Parakeet V3 (default, 25 European languages, 6.3% average WER) and V2, Cohere Transcribe (14 languages, 5.4%), OpenAI Whisper Small, Large V3 Turbo and Large (99 languages). All run as CoreML models on the Mac; any WhisperKit-format model can be added.
PrivacyTranscription and formatting run on the Mac. The app makes four network calls, none carrying audio or text: a model download, licence activation, a daily update check (optional) and feedback the user sends. No telemetry.
Claude CodeA hook (also a Claude Code plugin) shows Claude Code's plan, question and permission prompts as cards the user can answer by voice. Permission cards are click-only by design.
First public release1.2.0, August 31, 2026. Current version: 1.7.5.
Websitehttps://freeflow.audio
Contactsupport@freeflow.audio

Screenshots

Current build, light appearance, 1600 × 1007 PNG. Full-resolution 2× versions of the Models and Settings pages are linked below each image.

FreeFlow Dictation's Models page listing Parakeet V3, Parakeet V2, Cohere Transcribe, Whisper Small, Whisper Large Turbo and Whisper Large with size, languages and word error rate
Models: six on-device speech models with their size, languages and average word error rate. Download · 2× version
FreeFlow Dictation's Settings page showing the Hotkeys section: hold Option to talk, the hands-free hotkey, the conflict check, read selection aloud and copy last transcript
Settings: hold ⌥ to talk, a hands-free hotkey, read-aloud and Copy Last Transcript. Download · 2× version
FreeFlow Dictation's Modes page with the Voice to Text mode selected and a prompt to pick or create a mode
Modes: each mode keeps its own model, language, filler rules and formatting prompt. Download
FreeFlow Dictation's Dictionary page with replacement and vocabulary fields
Dictionary: replacements and vocabulary, kept on the Mac. Download

Use

Everything on this page may be used in articles, reviews, listings and videos about FreeFlow Dictation without asking, including the screenshots and the icon. Please do not alter the icon or imply endorsement. For a review licence, an interview, or screenshots of a specific feature, write to support@freeflow.audio; a person answers, usually within a day.

Download now