Speak.
We Type.
Any App.

Free native voice dictation for Windows, macOS, and Linux. Run locally and offline, or add optional cloud accuracy and API access.

No account required for the desktop app. Local inference stays on your machine with no usage meter.

QUICK_START.EXE
  1. 1Download the free app for Windows, macOS, or Linux. No account.
  2. 2Press Ctrl+Shift+D and speak.
  3. 3Your words are typed into whatever app has focus.
OR TRY IT IN YOUR BROWSER
Mac OS•Windows•Linux•REST API•Multilingual Cloud•Offline Mode•Apache-2.0•Audio Never Stored

Free Language Games

Train your ear, then train your mouth. DictatorFlow now includes browser games for listening practice and pronunciation feedback.

Pronunciation Practice

Speak a foreign-language phrase while your live waveform overlays the native target. Green bars mean your rhythm and stress match.

Practice with waveform scoring

Guess the Language

Hear a native-speed sentence, choose from four languages, and learn the answer with a translation after every round.

Play the listening quiz

Measured,
Not Marketed.

Every speed change is gated on word error rate against the same recordings, and we publish the numbers, including the ones that did not ship.

Local Mode

Free. Offline. Audio stays on your machine.

  • Parakeet CTC 0.6B on ONNX Runtime, downloaded from inside the app
  • NVIDIA CUDA and AMD ROCm builds, with CPU fallback
  • Experimental VibeVoice-ASR CPU path, opt-in only (60.45% WER on our nine-clip test, so not the default)

Cloud Mode

Optional. Routed with automatic fallback.

  • Requests route across several speech providers and fail over when one is slow or down
  • Multilingual transcription with no local model to manage
  • Audio is transcribed and discarded, never stored
How we optimize speed without sacrificing WERFast local transcription with remote fallbackVibeVoice-ASR benchmark report

Cross Platform

Native binaries for macOS Intel, Windows, and Linux x86_64. No Electron bloat. Written in Zig.

Many Languages

Cloud mode detects the spoken language automatically. The default local model is English-first.

Local / Offline

Run free local models on CPU, NVIDIA CUDA, or AMD ROCm builds. Your audio stays on your machine.

New

Edit With
Your Voice.

Select text, speak a command, and watch it transform. Rewrite paragraphs, fix code, translate, summarize -- all by voice. No typing required.

  • Select any text in any app
  • Speak your editing command
  • Result replaces your selection
  • Works in every text editor, IDE, browser
Selected Text
The quick brown fox jumps over the lazy dog. This sentence contains every letter of the alphabet.
Voice Command
"Make it sound more professional and formal"
Result
The agile auburn fox vaults gracefully over the indolent canine. This pangram encompasses every character of the English alphabet.

Try It Now.

No account needed. Just hit record.

Developer API

Integrate our engine from the backend, the browser, or an embeddable voice widget. Same transcription engine, whichever surface you need.

  • Multi-provider Fallback Chain
  • PCM, WAV, WebM, MP3, OGG Support
  • Bring Your Own Provider Key
  • Open browser widget with waveform modal + textarea mounting
  • Usage-based Billing ($0.004/sec)
VIEW FULL DOCSDOWNLOAD WIDGET
Browser Widget

Put a mic beside any textarea.

The widget opens a self-contained speech modal with a live waveform, Enter-to-finish, X-or-Esc cancel, and direct transcript insertion into inputs, textareas, or contenteditable surfaces.

Integration

API Quick Start

Pick the stack you already use. The payload is just raw audio bytes plus a bearer token.

cURL example
curl -X POST https://dictatorflow.com/transcribe \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: audio/wav" \
  --data-binary @audio.wav
JSON response
{
  "text": "Speak. We type. No errors.",
  "duration_seconds": 0.142
}
Live Widget Demo

The textbox you can talk into.

Add the Dictator widget to your site and let users convert speech to text in any input field.

Widget Docs

Free Local. Pay for Cloud.

Downloading the native app never requires a subscription. Charges apply only when you choose DictatorFlow's hosted models or developer API.

Local Desktop
$0 forever

Native desktop app and local inference with no account, subscription, or usage meter.

  • Mac, Windows, and Linux binaries
  • Unlimited local transcription
  • CPU, NVIDIA, and AMD paths
  • Continuous updates
Cloud Pro
$9/mo

Optional hosted transcription with the highest-accuracy managed models. Cancel anytime.

  • 10 hours/mo cloud transcription
  • No local model setup
  • Managed provider fallback
  • Free local mode still included
API Credits
$0.004/sec

Pay as you go. Top up credits anytime. No commitment.

  • REST API with API keys
  • Billed per second of audio
  • Managed provider fallback
  • Voice command endpoint
Open source

DictatorFlow Is Open Source.

The desktop app, the local speech engine, and the API server are released under the Apache License 2.0, patent grant included. Read it, self-host it, or fork it. Hosted cloud transcription and the developer API stay paid services, and that is what funds the free local build.

  • Apache-2.0 source, public mirror on app.nz
  • Local inference runs with no account
  • Audio is never stored server-side

The public mirror on app.nz goes live with the first tagged release.

Pump.fun

Dictator token is live.

Open listing