You speak. Text appears. 5× faster.
Local voice model: dictation plus AI Prompt and Agent Mode in one. Up to 5× faster than cloud tools.
Your keyboard eats your time. Cloud dictation eats your data.
You speak 3× faster than you type. Yet you still pound keys for hours every day. Why?
Email, Slack, notes, code comments. Hours of hammering keys instead of thinking or building.
Wispr Flow and friends stream every voice note to someone elses servers. Private conversations. Client names. All gone.
Cloud dictation lags. Local tools are clunky. By the time it reacts, you typed the sentence yourself.
Three modes. One key. Zero typing.
Nothing to relearn. Hold a key, speak, let go.
Three modes — Dictation, Prompt, Agent
Each mode has its own hotkey.
Click a tab to pause
Speech is transcribed and inserted as-is. No AI cleanup. 100% local.
“hey luke the 10am doesn't work can we move to 11”
Hey Luke, the 10am doesn't work. Can we move to 11?
Raw prompt in, structured prompt out.
You speak like a human. Voiceit’s Prompt mode reshapes it into the format AIs demonstrably understand better — with role, task, context, format.
- · No role setup
- · No style anchor
- · No format
- · AI guesses what you mean
- ✓ Role sets tone & expertise
- ✓ Clear task, no guessing
- ✓ Style anchor → consistent image
- ✓ Format = ready to use
"Think of Claude as a brilliant but new employee who lacks context on your norms and workflows. The more precisely you explain what you want, the better the result."
Anthropic recommends: set a role · clear task · format anchor — which is exactly what Voiceit’s Prompt mode does to your voice note. You speak freely. Structuring happens inside the app.
Source docs.anthropic.comNo tab-switch to ChatGPT to polish the prompt. No "wait, let me structure that." Hold ⌘ right, speak freely, done.
More than a dictate button. A full voice toolchain.
Eight things Wispr cannot do — or only halfway. Concrete, no marketing speak.
Audio stays local
Local voice model runs on-device. Your microphone signal never leaves your Mac. Wispr and Glaido stream it. We don’t.
3 modes instead of 1
Dictation types 1:1. Prompt shapes your speech into structured output. Agent transforms selected text. Wispr only has mode 1.
Clean output
No "uhms", no slips, no repetition. Reads like writing, not like speaking.
Custom profiles
Your own system prompts per mode. "Image prompt generator", "email reply", "code comment" — you create them, you trigger them.
Cross-app injection
Wherever your cursor is, the text lands. Mail, Slack, Notion, VS Code, Terminal. Works in any Mac app with a text field.
Custom hotkeys
Default is ⌘ left / right and ⌥ right. Bend it to your hand: Fn, Ctrl, any modifier combo.
Phrases · voice snippets
Speak a trigger → snippet lands. "my sig" becomes your email signature, "pitch intro" your standard opener. Up to 200 phrases per account.
Offline + history
Works without internet (dictation). Local history in SQLite, 3-day auto-cleanup. 7-day offline grace for Pro/Premium.
Faster. More private. Cheaper.
What you get — in numbers, not marketing speak.
150 words/min spoken vs. 40 typed
Local voice model runs on-device, no cloud round-trip
Wispr streams every word. Voiceit, not one
Dictation + Prompt + Agent in one app
Wispr charges from day 1. Voiceit, free forever
Sources · standard speak-vs-type speed · Apple Silicon benchmarks · in-house measurements
I use it myself. Every day.
Real stats from my own Voiceit install. Straight out of the local history DB. Today, 2026-05-14.
Today. On my own Mac.
Mail, Slack, code, notes.
At 172 words per minute.
At a 40 wpm typing pace.
Source · ~/Library/Application Support/com.newways.voiceit/history.db · Marcel Porcher · 2026-05-14
What you get for €15 a month — stacked.
No upsell. No add-on store. All in the Pro plan.
Custom profiles
Worth · €19 / monthYour own system prompts per mode. Email reply, code refactor, image prompt — triggered separately.
Phrases · voice snippets
Worth · €9 / month200 custom triggers. "my sig" → email signature. "pitch intro" → standard opener.
Custom hotkeys
Worth · €5 / monthDefault ⌘ left/right, ⌥ right. Bend it to your hand. Fn, Ctrl, any combo.
Direct builder access
Worth · pricelessBug? Email Marcel. Feature wish? Discussed in the next sprint. No ticket system.
EU GDPR · local voice model
Worth · no audio cloud cost — by designAudio stays on your Mac. Local transcription, nothing to the cloud. EU backend for auth and billing.
All features active in the Pro plan · no add-ons · no hidden pricing
Wispr dictates. Voiceit works.
Same price. More function. Audio stays on your Mac. Period.
As of 05/2026 · public info & in-house testing
With Wispr you pay for dictation. Here, dictation is free.
And for the same price as your Wispr subscription you get Prompt and Agent mode on top. Audio stays on your Mac.
Get VoiceitDictation is free. Agent is cheaper than Wispr.
No demo limit. No credit card to download. Start speaking, decide later if you want Agent. AI calls are pass-through: you pay volume, not markup.
Free
FreeDictation free & local. AI features (EU, no training).
- Dictation mode local (4,000 words/week)
- Dictation 100% local, audio stays on your Mac
- 20 AI calls / month (EU servers, Frankfurt)
- Up to 2 min recording
Pro
For daily voice users — emails, Slack, notes, prompts.
- ~16 AI calls/day (500/month) — a pitch, a meeting recap, a few emails
- Everything in Free
- Agent mode (custom prompts)
- Up to 8 min recording
- Priority on new features
Premium
For people who dictate every meeting and prompt every workflow.
- ~100 AI calls/day (3,000/month) — every meeting transcribed + transformed
- Everything in Pro
- Up to 20 min recording
- EU-hosted AI (GDPR-compliant)
- Early access to new models
Billed every 4 weeks · Stripe checkout · cancel any time
GDPR-compliant · Dictation local · AI in EU (Germany) · no AI trainingFrequently asked questions.
What people ask before they install.
Yes. Voiceit runs natively on Apple Silicon — M1, M2, M3, M4. macOS 13 or newer. Intel Macs are not supported because the local voice model is tuned for Apple Silicon.
Another question? Write directly to marcel@newways.ai.
Voiceit. Done.
Two-minute install. Learn one key. Typing is over. Dictation costs nothing and stays free.
Beta · GitHub Releases · built by newways.ai