REC · v3.1 · Released today // On-device · 27 languages · Sub-20ms

Speak.
It's already typed.

EcoScribe is a voice-to-text engine that lives in your menu bar. Hold a key, say what you mean, and watch words appear — anywhere on your Mac, with surgeon-grade accuracy and on-device privacy.

REC · Live transcript
WPM 214 WER 1.4% EN-US
Hold Space anywhere, anytime
Latency 18msAudio 48kHz
// Engineering by the numbers · updated weekly
18ms
End-to-end latencyFrom breath to text · p50, M3 Pro
1.4%
Word error rateEnglish · LibriSpeech clean benchmark
2.4M
Words / dayTranscribed on-device · last 30 days
0kb
Audio uploadedYour voice never touches our servers
02 · Features

Built for the way you actually think.

A focused tool, not a kitchen sink. Six things, done exceptionally — no plugins, no model picker, no cloud round-trip.

01 · Global Hotkey
+Spaceany app

One shortcut. Every text field.

Cursor in Mail, Slack, Linear, Xcode, Terminal — doesn't matter. Hold the key, speak, release. Words land where your caret is.

02 · On-Device AI
0ms
Cloud round-tripYour voice never leaves the device.

Private by physics, not policy.

Whisper-grade models run natively on Apple Silicon. No upload, no transcripts in someone else's logs. Works offline at 30,000 ft.

03 · Latency

Sub-20ms.

Words appear faster than you'd typed them.

04 · Smart Rewrite
"uhh send him the
file by friday"
→ "I'll send the document by Friday."

Filler & ums, gone.

Optional cleanup turns a ramble into a draft.

05 · Multilingual
ENESFRDEJAZH+21

27 languages.

Switch mid-sentence. We catch it.

06 · Custom Vocabulary
Kubernetes Vercel @priya.chen EcoScribe PostgreSQL Q3-OKRs

Knows your jargon, names, and acronyms.

Drop a list of teammates, products, code identifiers — EcoScribe learns to spell them right the first time. No more "Curt" instead of "Kubernetes."

07 · Modes
Email — polite, signed Code comment — terse Standup — bullets Journal — verbatim

One voice, four tones.

Pick a mode per shortcut.

08 · Keyboard-first
⌃SpaceRecord
⌃RRewrite
⌃MSwitch mode
⌃ZUndo

Every action has a shortcut. Mouse optional.

Designed for keyboard-first developers, writers, and operators. Customize every binding, including the recording finger.

03 · How it works

From breath to text. In three steps.

STEP · 01
+Space

Hold the hotkey.

A small red dot appears in your menu bar. You can start speaking before it finishes blinking.

STEP · 02

Just talk.

EcoScribe captures 48kHz audio on-device. Pauses, breaths, "um" — handled.

STEP · 03
It's typed. Right where your cursor was.

Release. Done.

The transcription is in your text field. No popups, no copy-paste, no review modal.

04 · The Library

Every word, searchable.
Auto-tagged, never uploaded.

EcoScribe keeps a local-only history of everything you dictated. Search "the part about Q3", drag a take into another app, or export the transcript as Markdown.

  • Full-text search — across every transcript, instantly
  • Auto-tagged — by app, time of day, language, mode
  • Markdown export — or copy as code comment, JSON, plain
  • Encrypted at rest — file-vault layer on top of local store
EcoScribe — Library
2026-05-16 · 09:41 · 1m 04s · MODE: STANDUP

Q3 planning brief

Ship the on-device translation beta by week three, then move the team onto reliability work. Add Priya to the loop on auth refactor — she has the most context on the new SAML edge cases. Keep the sync to fifteen minutes.
● standupen-usLinear1m 04s
05 · Pricing

One price. No tokens, no tiers of tokens.

Pay for the app, not the API behind it. Free forever for casual users; Pro unlocks the modes and the library.

Free
For trying it out and casual notes. Stay forever if it fits.
$0 forever
  • Unlimited dictation
  • Global hotkey
  • On-device English
  • 3 languages
  • 7-day history
Download free →
Team
Shared vocabularies, billing, and SSO. For studios & squads.
$14 / seat · billed yearly
  • Everything in Pro
  • Shared vocabularies
  • Centralized billing
  • SSO (Okta, Google, MS)
  • Audit log & admin console
  • Priority support
Talk to sales →
06 · FAQ

Things people actually ask.

01Does my voice ever leave my Mac?

No. Models run on-device using Apple's Neural Engine. The only thing that hits our servers is your license check (once a day) and crash reports — opt-in.

02Which Macs are supported?

Any Apple Silicon Mac on macOS 14 (Sonoma) or later. Intel Macs are not supported — the on-device models need the Neural Engine.

03How does it compare to dictation built into macOS?

Faster (sub-20ms vs ~1s), more accurate on technical vocabulary, and it works in every text field — including Terminal, Xcode, and SSH sessions.

04Can I cancel anytime?

Yes. Pro is monthly or yearly, no contract. If you cancel, your Library stays on disk forever — we don't lock data behind a subscription.

05Is there a Windows or Linux version?

Not yet. A Windows beta is planned for Q4 2026. Sign up on the changelog page to be notified.

06What about students or open source?

Free Pro license for verified students, and lifetime Pro for committed open-source maintainers. Email hello@ecoscribe.app.

Stop typing.
Start saying it.

Download EcoScribe for Mac