I built HyperVoice because typing was the slowest part of my day. Press one key, talk, and your words land at the cursor in any app — voice to text with the speech engine running locally on your machine, so your audio never leaves it.
Local speech engine — audio never leaves your machine
Sub-second with GPU acceleration
🪟 Windows 10+
·🐧 Linux x64 Beta
·🍎macOS 11+ Beta
Already powering real dictation
120,000+
Words dictated
6,100+
Dictations transcribed
810+
Accounts created
How It Works
Hotkey. Speak. Done.
The whole product is one loop, about three seconds long. Here it is, step by step.
HyperVoice
The floating pill — this is the actual animation from the app: orb while you speak, amber while it transcribes, green when your words have landed.
01
Press the hotkey
Ctrl+Shift+Space from any app — or rebind it. Tap to toggle, or hold to push-to-talk. The app sits in your tray; you never alt-tab to it.
Ctrl+Shift+Space
02
Say the thing
Whisper or Parakeet transcribes on your machine — your audio never leaves it. With a GPU (Vulkan: NVIDIA, AMD, Intel) it's back in under a second.
03
Words land at the cursor
The text pastes itself into whatever was focused — editor, email, terminal, chat. Want it cleaned up first? Optional AI styles rewrite it on the way through.
500 words/day free forever. No credit card required.
The Math
Speaking Is ~4x Faster Than Typing
I type about 40 words a minute on a good day. I talk at 150. That gap is the whole reason I built this.
40
WPM typing
Average typing speed
150
WPM speaking
Average speaking speed
3.75x
Faster input
Voice vs. keyboard
2 hrs
Saved per day
If you type 3 hours daily
Typing40 WPM
Speaking with HyperVoice150 WPM
A 1,000-word email takes 25 minutes to type but only ~7 minutes to dictate.
Core Features
What's in the box
Everything here ships in the free download. The speech engine runs on your machine; the optional AI cleanup is the only cloud piece.
Works in Any App
Press Ctrl+Shift+Space, talk, and the words land at your cursor — VS Code, Slack, Gmail, a terminal, anything with a text field.
On-Device Speech Engine
Whisper and Parakeet run on your machine — your audio never leaves it, and transcription works with no internet at all.
GPU Acceleration
Vulkan means any GPU — NVIDIA, AMD, Intel — on Windows and Linux. Sub-second results; falls back to CPU if there's no GPU.
The Floating Pill
A tiny always-on-top pill shows what's happening — orb while you speak, timer, green flash when the text lands. Drag it anywhere, or hide it.
Custom AI Styles
Optional cloud cleanup: speak rough, land polished. 7 built-in styles plus your own prompts — via your API keys or HyperVoice Cloud on Pro.
History, Stats & Dictionary
Every dictation is searchable locally, your stats show how much typing you've skipped, and a custom dictionary handles jargon and names.
Works Everywhere
Talk Instead of Type
If you can type in it, you can talk into it — I dictate into Claude Code, Slack, and email all day. It's also a genuine relief if RSI or hand strain makes typing the painful part.
AI Chat
ChatGPT, Claude, Gemini
Teams & Chat
Slack, Teams, Discord
Code Editors
Cursor, VS Code, Claude Code
Anywhere
Email, browser, docs, any app
System-level text injection — HyperVoice works with every application on your computer.
Optional — off by default
Speak Rough, Land Polished
By default you get the raw transcript. Flip on an AI style and a model rewrites it before it lands — ums removed, email formatted, ticket structured. This step runs in the cloud: your own OpenAI or Anthropic key, or HyperVoice Cloud on Pro.
Clean Up
Remove filler words, fix grammar, and tighten your speech into clean written text.
Professional Email
Speak your thoughts casually, get a polished professional email ready to send.
Chat Message
Speak naturally, get a casual but professional message ready for Teams or Slack.
Meeting Notes
Ramble about a meeting for 2 minutes — get structured notes with action items.
Bullet Points
Convert a stream of consciousness into a concise, organized bullet-point list.
Status Update
Describe your day's work and get a formatted status update with progress and next steps.
Ticket / Issue
Describe a bug or task verbally and get a structured ticket ready for Jira or GitHub.
Plus unlimited custom styles — write your own system prompt once and it becomes a one-tap mode.
<1s
Transcription latency (with GPU)
11
Speech models
99
Languages supported
100%
On-device transcription
Your audio never leaves your machine.
Why HyperVoice
Where I planted the flag
Most dictation tools are macOS-only, cloud-only, or subscription-only. I made the opposite calls on all three.
Cross-platform
Vulkan GPU Acceleration
One of the very few dictation apps with GPU acceleration on both Windows and Linux. Most alternatives run CPU-only or are macOS-only.
Accurate
99 Languages Supported
Transcribe speech in 99 languages with state-of-the-art accuracy. Multiple model sizes to match your needs.
Pay once
Lifetime License — No Subscription
Buy once and own it forever, while most dictation tools are subscription-only. Prefer to subscribe? Pro starts at $6.67/mo — or bring your own API keys for zero-markup AI cleanup.
Lightweight
Native Performance
Tiny memory footprint, instant startup. No bloated runtimes. Debug panel shows all pipeline timings.
Flexible
Configurable Everything
Custom hotkeys, push-to-talk or toggle mode, 11 AI models, GPU or CPU mode, dark/light theme, auto-paste toggle. Make it work the way you want.
Private
Your Data Stays Yours
Audio and transcription stay on-device; history is stored locally. We log only technical metadata (model, latency, errors) to keep the service healthy — never your audio or transcribed text. Cloud processing is entirely opt-in.
How We Compare
Local engine, cross-platform, pay once
An honest look — the alternatives win some rows too. Full breakdowns on the comparison pages.
HyperVoice
Wispr Flow
Superwhisper
Dragon
Where it runs
On your device
Cloud servers
On your device
On your device
Platforms
Windows, Linux, macOS
Windows, Mac, mobile
macOS only
Windows only
GPU acceleration
Vulkan — any GPU
Cloud-side
Metal — Apple only
CPU only
Pay once (lifetime)
Yes — $49.99
No — subscription
One-time option
Yes — $200–700
Voice training
Not needed
Not needed
Not needed
Required
Free tier
500 words/day
None
Limited
Trial only
Competitor details from our comparison pages, current as of 2026. Products change — see each vendor for the latest.
Pricing
Start free. Pay if it sticks.
500 words a day, free forever — enough to know within a week whether this is for you. Upgrade when the cap pinches or you want the built-in AI cleanup.
I kept a true lifetime license: $49.99 once, yours forever. Most dictation tools won't sell you that.
I build HyperVoice in public — much of it live on camera in the “Vibe Coding” series on YouTube. Here's what I'm working toward next.
Cloud STT
Cloud-based speech-to-text for higher accuracy on lower-end hardware.
macOS & iOS
Linux shipped (it's in beta now) — Apple platforms are next on the list.
Built-in Translation
Dictation already works in 99 languages — next up: speak in one language, paste in another.
Meeting Recording
Record and transcribe meetings with speaker diarization.
File Transcription
Drag and drop audio/video files for batch transcription.
FAQ
Things people actually ask
Does HyperVoice send my audio to the cloud?
No. Speech-to-text runs entirely on your device — your audio never leaves your machine. The optional AI cleanup step is the one cloud piece: it sends the transcribed text (never audio) to your own OpenAI/Anthropic key, or through HyperVoice Cloud on Pro. Leave it off and nothing is sent anywhere.
Is my data stored when I use cloud post-processing?
No. Your text is sent to the AI provider, processed in real-time, and returned — nothing is saved on our servers. We log basic usage metadata (like which model was used and token counts) to monitor service health, but your actual text content is never stored or retained.
Can I use HyperVoice for coding?
Yes! HyperVoice injects text at your cursor in any application, including Cursor, VS Code, and CLI tools like Claude Code. Dictate comments, documentation, commit messages, or chat with AI coding agents.
What GPUs are supported?
GPU acceleration works with NVIDIA, AMD, and Intel graphics cards. If no compatible GPU is found, HyperVoice automatically falls back to CPU-based transcription.
What AI models are available?
Eleven Whisper models from Tiny (75 MB) to Large-v3 (3.1 GB) — larger models are more accurate but need more VRAM. HyperVoice also includes NVIDIA Parakeet as an alternative engine, which is especially fast on CPU. Choose the right balance of speed and accuracy for your hardware.
Can I use HyperVoice offline?
Yes — transcription works completely offline once you've downloaded a speech model. The only feature that needs internet is the optional AI cleanup, because that step runs in the cloud.
Windows shows a security warning when I install. Is this safe?
Yes, HyperVoice is safe. As a new application, Windows SmartScreen may display a “Windows protected your PC” warning. Simply click More info then Run anyway to proceed. This is standard for new software and will go away as more users install the app. Code signing is coming soon.
What platforms are supported?
Windows 10+, Linux x64 (beta), and macOS 11+ on Apple Silicon (beta). The Linux build ships as a single AppImage and needs an X11 session; the Mac build is an Apple Silicon (M-series) .dmg. iOS is still on the roadmap.
Does it handle accents and non-native English?
Yes. HyperVoice runs OpenAI's Whisper, which was trained on a huge range of speakers and accents, so it handles non-native and accented English well. Accuracy improves with the larger models — and it transcribes 99 languages, not just English.
Can I dictate hands-free? Toggle or push-to-talk?
Both. Choose toggle mode (tap your hotkey to start, tap again to stop) or push-to-talk (hold the hotkey while you speak). HyperVoice can also launch at startup and sit quietly in your system tray, so it's always one keystroke away.
What are the system requirements?
HyperVoice is a small native app with a tiny memory footprint. You'll need disk space for whichever model you download (from 75 MB for Tiny up to 3.1 GB for Large-v3) and enough RAM or VRAM to run it. A GPU is optional — it makes transcription sub-second — but everything also runs on CPU.
Does it work on Wayland?
The Linux build currently needs an X11 (Xorg) session. If you're on Wayland, you can pick “Xorg” or an X11 session from your login screen's gear menu. Native Wayland support is on the roadmap.
Can HyperVoice help with RSI or accessibility?
That's one of the most common reasons people use it. Dictating instead of typing takes the load off your hands, which helps with RSI, hand strain, and limited mobility — and because text is injected at your cursor in any app, it works across your whole workflow, not just one editor.
Is it “HyperVoice” or “hyper voice”?
It's HyperVoice — one word. A lot of people type it as two words (“hyper voice”) or search “hypervoice ai” — that's all us. HyperVoice is the dictation hotkey for Windows, Linux, and macOS: press a key, speak, and your words appear at your cursor in any app, with speech-to-text running locally on your machine.
How does HyperVoice compare to Wispr Flow, Superwhisper, or Dragon?
We have honest side-by-side comparisons at /vs. Short version: vs Wispr Flow — both have great UX, but HyperVoice runs Whisper locally on your machine so your audio never reaches a cloud. vs Superwhisper — same architecture (local Whisper, GPU-accelerated), but HyperVoice is cross-platform across Windows, Linux, and macOS, while Superwhisper is macOS-only. vs Dragon NaturallySpeaking — HyperVoice uses modern Whisper AI (no voice training needed) at a fraction of the price.
Who builds HyperVoice?
I do — Christopher Cousens, solo. Most of it gets written live on camera in the “Vibe Coding” series on YouTube, bugs and all. If something's broken, tell me on the feedback page or Discord — fixes usually ship within days.
How much does HyperVoice cost?
HyperVoice has a free tier with 500 words per day — no credit card, no time limit. For unlimited usage, choose a Lifetime license at $49.99 (one-time, yours forever) with local transcription, GPU acceleration, 11 AI models, and BYOK support; or Pro at $7.99/month ($79.99/year) which adds HyperVoice Cloud AI processing. Paid plans include a 7-day free trial. Cancel anytime.
Ready to stop typing?
Download it, press Ctrl+Shift+Space, and say something. That's the whole onboarding.
