Marlowe

Mac, Windows, and the browser

As fluid as Wispr Flow. Four dollars a month.

Hold a key, talk the way you actually talk, and Marlowe puts finished text at your cursor. People search for this as WhisperFlow. The product is Wispr Flow, and it is excellent. Marlowe is the same motion — for a quarter of the price, on an engine you can actually read.

Ctrl+Spacehold, speak, release
Monthly
$4
Yearly
$3
Free cap
2k

Words a week on the free tier. Yearly is $36, paid once.

Priya

Can you push the launch note before tomorrow?

Hold the button, or Ctrl + Space. A short tap locks until you tap again.

Names it should spell

This preview transcribes with a cloud speech model, then runs the Marlowe writing pass. The Mac and Windows apps can run the same pass on Whisper, on the machine.

From VoiceStudio

Transcriptions

The Electron workspace from VoiceStudio: dictate or upload, keep the raw transcript and the refined line apart, and store the newest 200. Timings show up when the engine returns them. VoiceStudio is AGPL-3.0 — this is Marlowe’s own screen of that workflow, not a copy of their app.

Transcribe

Nothing saved yet. Start dictation or upload audio.

Dictionary words in the keyboard (none yet) are sent with each take.

History

0 / 200

History stays empty until the first transcript.

Built to stand next to Wispr Flow

Wispr Flow made dictation feel finished. Marlowe copies that behavior on purpose: the pause, the cleanup, the paste. It does not copy their invoice. This is a comparison, not an affiliation — Wispr Flow is their trademark.

Price

Marlowe · $4 a month, or $3 billed yearly

Wispr Flow · About $15 a month after a small free cap

How you talk

Marlowe · Hold a key, speak, release

Wispr Flow · Hold a key, speak, release

What lands

Marlowe · Finished writing, not a raw transcript

Wispr Flow · Finished writing, not a raw transcript

Where

Marlowe · Mac, Windows, and this browser

Wispr Flow · Mac, Windows, and iPhone

Engine

Marlowe · Open Whisper on the desktop apps. Cloud speech in this preview.

Wispr Flow · Proprietary cloud

Dictionary

Marlowe · Your names are biased into the recognizer

Wispr Flow · Yes

Audio on the machine

Marlowe · Yes, in the Mac and Windows apps

Wispr Flow · No — audio goes to their cloud

Heard

um hey so I was thinking we should push the launch to Thursday if that's okay uh I can send the notes tonight

Wrote

Hey — I think we should push the launch to Thursday. I can send the notes tonight.

Five open-source projects, one keyboard

Marlowe does not pretend to have invented speech recognition. Handy is the push-to-talk shell, Whisper is the accuracy target, and VoiceStudio is the transcription workspace — dictate, upload, history, and handoff.

  1. You speak
  2. whisper.cpp or faster-whisper
  3. Writing pass
  4. Cursor

01

Whisper

The accuracy target. Large-v3 is the open model the desktop apps are built to run — punctuation, casing, and about ninety-nine languages from the same weights.

MITopenai/whisper

02

whisper.cpp

The Apple and Windows CPU runtime. Metal on Apple silicon, AVX on Intel and AMD, a small native binary. This is what a Mac menu-bar build actually calls.

MITggml-org/whisper.cpp

03

faster-whisper

The NVIDIA Windows runtime. CTranslate2, int8, often about four times the original PyTorch Whisper at the same transcript. Used when a Windows PC has an NVIDIA GPU.

MITSYSTRAN/faster-whisper

04

Handy

The dictation shell. Push-to-talk, a history of takes, and paste-at-the-cursor — the interaction people already trust from the best open-source desktop dictation app.

MITcjpais/Handy

05

VoiceStudio

The transcription workspace this page uses: dictate or upload, raw and refined text kept apart, newest-first history of 200, timings, copy, export, and handoff into the keyboard. Their Electron app is the local desktop. Marlowe does not relicense it.

AGPL-3.0debpalash/VoiceStudio

Whisper, whisper.cpp, faster-whisper, and Handy stay MIT. VoiceStudio is AGPL-3.0: Marlowe links it and follows its transcription workflow in original code. It does not ship their Electron binary or relicense their source. The writing pass — filler stripped, tone matched, names kept — is the part priced at four dollars. In this browser the speech step is cloud, because those local engines are multi-gigabyte downloads.

Mac, Windows, and the VoiceStudio desktop

The local app is VoiceStudio’s Electron build. It runs on your machine. This preview cannot install it for you — these are their install commands, unchanged.

Mac and Linux

# Latest Electron release
curl -fsSL https://voicestudio.sh/install | sh

# A specific published Electron release (replace X.Y.Z)
curl -fsSL https://voicestudio.sh/install | sh -s -- --version X.Y.Z

# Build current main and install the desktop app
curl -fsSL https://voicestudio.sh/install | sh -s -- --main

# Uninstall the app, keeping your data
curl -fsSL https://voicestudio.sh/install | sh -s -- --uninstall

Windows uses the Electron EXE from debpalash/VoiceStudio releases. PowerShell can also run irm https://voicestudio.sh/install | iex.

Marlowe9:41

Menu bar · hold Option

Release to paste into the frontmost app.

  • whisper.cpp with Metal on Apple silicon
  • Whisper large-v3 turbo while on battery, large-v3 when plugged in
  • Right Option is the hotkey. It never touches the key you use to type.

System tray · hold Ctrl + Win

Paste with the same writing pass.

  • faster-whisper int8 when an NVIDIA GPU is present
  • whisper.cpp when it isn’t — same weights, no CUDA required
  • History, dictionary, and tone travel with the app, not with an account.
WindowsTray

Three to five dollars. We picked four.

A heavy day of dictation is minutes of audio, not hours. Priced so a month of talking still fits inside a coffee.

Free

$0

  • 2,000 words a week
  • Literal and clean tones
  • Browser keyboard

Marlowe

$4 / month

  • Unlimited dictation for a real workday
  • Message, email, notes, and code
  • Personal dictionary
  • Mac menu bar and Windows tray, with on-device Whisper
Try the keyboard

Marlowe is independent. Wispr Flow is a trademark of its owner. Whisper, whisper.cpp, faster-whisper, and Handy remain MIT in their own repositories. VoiceStudio is AGPL-3.0 and remains theirs. This preview is the keyboard and the transcription workspace. It is not their desktop installer.