Voice to text for Mac with smart cleanup, live translation, meeting summaries, and voice notes. No cloud. No subscription. Your voice never leaves your Mac.
Download Free100 free dictations · 50 notes · 10 meeting recordings · $49 $24.50 with VEXT50
brew install muvon/tap/vextmacOS 14+ · Apple Silicon · No account required
Three steps. No setup. No account.
Press and hold your configured shortcut. Vext starts listening instantly.
Talk at normal speed in any app. Your voice is transcribed in real time, locally.
Let go. Text appears at your cursor — wherever you are. No copy-paste needed.
Dictate, record meetings, or capture quick notes — all processed locally on your Mac.
Hold a hotkey, speak, release — text appears at your cursor. Works in any app, any text field.
Record with speaker labels, get full transcripts, screenshots, and AI-generated summaries.
Quick voice remarks — transcribed, cleaned up, and stored locally in the app for later.
Vext automatically cleans up your speech — removes filler words, fixes structure, keeps your intent intact. Same language, same meaning, just polished. Works alongside translation in a single pass.
"so um I was thinking we should uh probably move the deadline to like next Friday because ah the team needs more time to finish the uh the integration tests"
"We should move the deadline to next Friday — the team needs more time to finish the integration tests."
No audio uploaded. No cloud transcription. No account. This isn't a privacy policy — it's architecture.
Whisper runs directly on your Apple Silicon GPU. All processing stays on-device.
Works offline, on an airplane, behind a firewall. No connection needed, ever.
No analytics. No telemetry. No accounts. We never see your audio or transcriptions.
100 free dictations, 50 notes, and 10 meeting recordings to start. Then one price, unlimited use.
Launch promo · code VEXT50 auto-applied · until June 1, 2026
$120–360/year
Unlock from within the app when you're ready. Free updates included. Major new versions at 50% off for existing owners.
Feature-by-feature against the leading voice and meeting tools.
| Vext | Wispr Flow $12–15/mo | Granola $14–35/mo | Otter.ai $8–17/mo | |
|---|---|---|---|---|
| Dictation (paste at cursor) | ||||
| Meeting transcription | ||||
| Voice notes | ||||
| Speaker labels | ||||
| Cross-meeting voice recognition | N/A | |||
| AI text cleanup | ||||
| Meeting summaries | ||||
| Live translation | ||||
| Screenshot capture (any mode) | ||||
| Screenshots auto-paste to AI | ||||
| YOLO mode (auto-submit) | ||||
| 100% local / private | ||||
| Works offline | ||||
| No bot joins your call | N/A | |||
| Cost after 2 years | $24.50 | $288–360 | $336–840 | $200–408 |
Competitor data sourced from public websites as of April 2026. Features and pricing may change.
Vext transcribes locally in real-time — no upload, no server, no spinner. Your words are already there.
60 seconds of audio transcribed in ~400ms. On-device.
Let's review the Q3 roadmap and figure out priorities.
I think we should focus on the API redesign first. It's blocking three other teams.
Agreed. Can we have a draft by end of next week?
Record any meeting — Zoom, Google Meet, FaceTime, or in-person — and get a full transcript with speaker identification. Turn on Summarize to extract key points and action items. Both versions are always saved.
Vext detects every distinct voice in your meeting automatically. Name them once — and from your next call onward, the same person is identified, labeled, and color-coded without lifting a finger.
Capture any region of your screen during hands-free dictation. The screenshot pastes alongside your transcribed prompt — straight into Claude Code, Cursor, or any AI tool. Fully hands-free coding.
Check the API rate limits before we push the new integration. Sarah mentioned the sandbox has different thresholds than production.
The onboarding flow needs a skip option on the second screen. Users are dropping off because they think the setup is mandatory.
Try SwiftUI NavigationSplitView for the sidebar instead of the custom implementation. It handles state restoration automatically.
Press a key, say what's on your mind, and move on. Vext transcribes, cleans up, and stores your note locally — ready when you need it.
Talk in English, get the text in Russian. Or Spanish. Or Japanese. Vext translates your speech in real time as it transcribes — so the text that lands at your cursor is already in the language you need.
"Let's schedule a meeting for next Tuesday to discuss the project roadmap and assign tasks to the team."
"Давайте назначим встречу на следующий вторник, чтобы обсудить план проекта и распределить задачи в команде."
Vext is a system-level service. It works in any text field, in any app — browsers, editors, terminals, email, chat.
Press a key once to start dictation. Press again to stop. No holding required — just talk as long as you need. Perfect for longer passages or when your hands are busy.
Capture screenshots mid-dictation — they auto-paste with your transcript. See how →
Turn it on and Vext automatically presses Return after pasting your transcription. Speak, release, and your prompt is already running.
Stop editing. Stop polishing. Just talk. LLMs know what you mean, even when your words aren't perfect.
When you start recording, Vext automatically fades your system audio so your voice comes through clearly. Release the hotkey and volume returns to normal. No manual adjustment needed.
Pick the speech and AI models that fit your workflow. Run locally by default, or bring your own API key.
| Model | Type | Speed | Size |
|---|---|---|---|
| Parakeet Default NVIDIA NeMo — fast, accurate, optimized for Apple Silicon. | Local | 150x RT | ~600 MB |
| Apple Dictation Built-in macOS speech recognition. Zero download. | Local | Realtime | Built-in |
| OpenAI-compatible Any OpenAI-compatible STT endpoint. Bring your own API key. | API | Varies | — |
| Model | Type | Size |
|---|---|---|
| Gemma 3 4B Default Default. Fast, accurate, great balance of speed and quality. | Local | ~2.5 GB |
| Gemma 3 1B Ultra-lightweight. Fastest local option, lower accuracy. | Local | ~1 GB |
| Qwen 3 4B Strong multilingual support. Good for translation tasks. | Local | ~2.5 GB |
| LLaMA 3.2 3B Meta LLaMA. Strong general-purpose performance. | Local | ~2.4 GB |
| Phi-3.5 Mini Microsoft Phi-3.5. Compact, strong reasoning. | Local | ~2.8 GB |
| OpenAI-compatible Any OpenAI-compatible API — GPT, Claude, Gemini, or self-hosted. | API | — |
Newsletter writer, YouTuber, indie hacker? We pay 40% on every sale you bring — automatic monthly payouts, 60-day cookie window, no caps.
Vext is built by Muvon Un Limited, a small product studio. We build tools we use every day — designed to work locally, respect your privacy, and get out of your way.