Native macOS dictation · Processed on-device
Hold a key. Say it the way you'd say it, umms, backtracks and all. ShoutFlow turns it into clean, ready-to-send text at your cursor, in any Mac app. And in local mode, your voice never leaves your Mac.
How it works
Anywhere on your Mac: your editor, your email, a text field in a browser. Hold to record or tap to toggle. ShoutFlow lives in the menu bar, so there's no window to find.
Filler words, restarts, "wait, no" corrections. Speak like you think. In local mode the floating overlay shows your words landing live as you talk.
Clean text, pasted at your cursor.
Release the key. The AI strips the umms, fixes grammar and self-corrections, formats for the app you're in, and pastes the result right where you were typing.
Context-aware tones
ShoutFlow detects the app in front and adjusts the writing to match. Casual in Slack, buttoned-up in Mail, term-exact in your editor. Same voice, right register, no manual switching.
"hey um can we move standup to nine thirty tomorrow, my train's delayed"
"uh tell them the quote's approved and um if the contract's signed by friday we can kick off monday"
"make fetch user async throws and uh rename it fetch current user, keep the retry logic"
PLUS OUTPUT STYLES · WHATSAPP MESSAGE · EMAIL REPLY · VIBE-CODING SPEC · WHEN YOU WANT A SPECIFIC SHAPE.
The two-stage engine
Dictation is two jobs: turning audio into words, and turning words into writing. ShoutFlow lets you choose each stage independently: fully on-device, or bring your own API key to the provider you trust. No middleman, no markup on tokens.
Mix anything with anything. Local + local for total privacy. Local + OpenRouter for frontier clean-up. xAI + DeepSeek because it's Tuesday. Switch per stage, in Settings, whenever.
Privacy
Most dictation apps route your voice through their servers and ask you to trust them. ShoutFlow is built so there's nothing to trust. In local mode, this is the entire path your voice takes.
Captured locally, with a visible recording indicator.
WhisperKit transcribes on the Neural Engine. An MLX language model cleans up on the GPU. Wi-Fi off? Still works.
Clean text pasted into the app you're using. That's the whole trip.
See your license status. Never your audio, your text, your prompts, or your API keys. By design, they're not in the path. Choose a BYOK provider and your Mac talks to it directly instead.
Transcription and clean-up both run on Apple Silicon after a one-time model download. Airplane mode is a supported configuration.
Provider credentials live in the macOS Keychain. Never in plaintext, never in logs, never on our servers.
Choose a cloud provider and your Mac calls it straight over TLS. ShoutFlow never proxies, logs, or stores what you say or write, and shows you exactly which stage sends what, where, before you enable it.
No account. No analytics SDK. License checks carry no audio, no text, no usage data, and keep working offline with a generous grace period.
The details
Teach it names, jargon and product terms once: Priya, Kubernetes, Recurse, spelled right every take.
Say "my email" and your address appears. Trigger phrases that expand into full, formatted text.
Watch words land in real time while you speak in local mode. No black box between you and the take.
Your last 50 transcriptions, one click from the menu bar. Copy any of them back out instantly.
A draggable floating console with record, pause, history and settings. Park it wherever you work.
WhatsApp message, email reply, or a structured vibe-coding spec. Choose the shape of the result.
Hold-to-record or press-to-toggle. Default ⌘⇧Space, remap to anything.
Swift 6, menu-bar only, no Electron, no dock clutter. Launches at login and stays out of your way.
Pricing
The popular dictation apps rent you your own voice for $12–15 a month, forever. ShoutFlow is a tool, and you buy tools once.
ONE-TIME PAYMENT VIA STRIPE · SOLD BY RECURSE LTD
MACOS 14+ · FULLY-LOCAL MODE REQUIRES APPLE SILICON · INTEL MACS RUN BYOK CLOUD PROCESSING
FAQ
Not for dictation. In local mode, transcription and clean-up run entirely on your Mac after a one-time model download. Airplane mode works. BYOK cloud modes talk to the provider you chose. A brief online check activates your license, then it keeps working offline through a generous grace period.
macOS 14 (Sonoma) or later. Apple Silicon unlocks the fully-local mode: WhisperKit on the Neural Engine, MLX clean-up on the GPU. Intel Macs are supported with BYOK cloud processing for both stages.
WhisperKit for speech-to-text and an MLX language model for clean-up. Models download on first use and you choose the size, from fast-and-light to large-and-meticulous. Nothing is bundled, nothing is uploaded.
If you want cloud models, you plug in your own API key: OpenAI, OpenRouter, DeepSeek, Gemini, Kimi, xAI, or a compatible custom endpoint. Requests go directly from your Mac to that provider, billed by them at cost. Your keys are stored in the macOS Keychain and never touch our servers.
Really. $25 once, via Stripe, and the app is yours on up to two Macs. If you use BYOK cloud providers, you pay them directly for exactly what you use. We take no cut and add no markup.
Yes. Download and try the full app before buying. No account and no card required.
Email support within 14 days of purchase and we'll refund you. No forms, no exit interview.
ShoutFlow is a direct download: Developer ID signed, notarized by Apple, and Gatekeeper-approved. It asks only for the permissions dictation needs (microphone, accessibility for pasting), and tells you exactly why.
Pay once. Speak everywhere. Keep your voice on your Mac.
MACOS 14+ · SIGNED & NOTARIZED · SOLD BY RECURSE LTD