Step 01
🎙
Record

Press your hotkey and talk naturally. No need to slow down or speak formally. Just say what you mean.

Step 02
Transcribe

AI on your Mac converts speech to text instantly. Everything runs locally. Your voice never leaves your device.

Step 03
Polish

Grammar is fixed, filler words are removed, and your meaning is preserved automatically.

Step 04
📋
Paste

Clean, polished text appears wherever you're typing, in any app, any field, without switching windows.

Everything happens in parallel

Recording, transcription, and polishing overlap. That's how EnviousWispr delivers results so fast.

Real-time pipeline
Streaming from voice to paste
What you said Streaming ASR
Pipeline
Record
5.0s
Stream ASR
real-time
parallel
Finalize
~80ms
Pre-warm
TLS
SSE Polish
~1.5s
Paste
<5ms
What got pasted SSE Streaming

Built to be fast and private

~2s
From voice to paste
End-to-end latency
2
Speech engines
WhisperKit + Parakeet
5
Polish engines
EG-1, Apple, Ollama, OpenAI, Gemini
0
Audio uploaded
Cloud polish sends text only by choice
Apps supported
Any text field on macOS

Choose how your text gets cleaned up

Turn cleanup on and pick your engine. Each one is good at something different, and every option except a cloud key runs on your Mac.

Start instantly

Apple's built-in AI

On the latest macOS, cleanup runs on-device with nothing extra to download. The zero-setup way to get started.

On-device · macOS 26+ · No download
Your call

Your own local model

Run any local model through Ollama, fully offline, if you would rather bring your own.

On-device · Bring your own model
Your call

Your OpenAI or Gemini key

Prefer a cloud model? Add your own key. Only your text is sent, never audio, and we ask the provider not to keep it.

Cloud · Your own key · Text only

How EG-1 compares on cleanup quality

Share of 1,890 real dictation cases each model cleaned up correctly, scored on the same cases by the same judge.

EG-1 on-device
93.7%
Gemini 3.5 Flash cloud
92.6%
GPT-5.4-mini cloud
83.8%

Our own benchmark, not an independent review. The eval harness and prompts are public; the test cases are personal dictations and stay private.

Everything you need, nothing you don't

🧠
On-device transcription

Powered by WhisperKit and Parakeet, transcription runs entirely on your Mac. No audio ever leaves your device.

✏️
Smart polish

Filler words removed, grammar corrected, your meaning preserved. Runs on-device with EG-1 or Apple Intelligence, or with your own OpenAI or Gemini key.

🎯
Voice activity detection

Automatically detects when you start and stop speaking, so you never have to babysit the recording process.

🖥
Works in any app

Polished text pastes into any macOS application: email, Slack, Notes, browser forms, terminals. Everything.

🔒
Push-to-talk and hands-free

Hold your hotkey to record, or double-press to lock into hands-free mode for longer dictation sessions.

🔌
Multiple AI providers

Use on-device polish with EG-1 (our own model), Apple Intelligence, or Ollama, or bring your own OpenAI or Gemini key. Cloud polish requests include store: false so providers are asked not to retain the text.

🔤
Custom word correction

Teach the app to recognize names, acronyms, and domain-specific terms that general models consistently get wrong.

🍎
Native macOS app

Built in Swift for macOS 14 and up. Lives in your menu bar, uses minimal resources, feels at home on your Mac.

⚙️
Streaming pipeline

Recording, transcription, and AI polishing run in parallel. The result arrives almost the moment you stop speaking.