Speech to text · Windows 10 + 11
Speech to text for Windows — free, offline, in any app
Press one hotkey, speak, and text appears at your cursor — in Word, Slack, Gmail, code editors, anywhere. Free local engine. Optional OpenAI Cloud tier for AI rewrites and web search.
Windows 10 or 11 (64-bit)
Speech to text on Windows, without the friction
Speech-to-text software on Windows used to mean training a profile for an hour, then dictating into a single application. Whisper by Remskill replaces that with a global hotkey: push it, talk, release. Your words land where your cursor is — in Outlook, in a comment field, in Cursor, in a Discord DM. Nothing to launch, nothing to click, nothing to retrain when you switch apps.
Two engines ship in the same installer. OpenAI's Whisper covers 99 languages and is the model that defined modern open-source speech recognition. NVIDIA's Parakeet is faster on CPU and runs 5–10× quicker than Whisper for English and 24 European languages. You choose per-task. Both run fully on your machine — no audio leaves the device unless you opt into Cloud mode.
Whisper Pro adds the OpenAI Cloud surface: GPT-class transcription, AI rewrites (formal, casual, bullet list, custom), and live web search by voice. It is optional and gated behind a hotkey; the free local mode keeps working without it.
Three keys, anywhere on Windows
One global hotkey. The same loop in every app from File Explorer to VS Code.
Press the hotkey
Default is the Right Option key, configurable to any chord on a Windows keyboard. A small overlay appears in the corner — you'll see the waveform live.
Speak naturally
No wake word, no training, no per-app setup. Talk in any of 99 languages; Whisper handles code switching mid-sentence.
Text pastes at your cursor
Release the key. Whisper transcribes locally (or via Cloud), optionally rewrites with AI, and types the result exactly where your cursor is. Works in every Windows app.
What you get on Windows
Built for typing-heavy work — writers, developers, support reps, students, anyone with a Windows machine and a microphone.
Global hotkey in every app
Push to talk in Word, Outlook, Notion, Slack, Discord, Chrome address bar, any field that accepts text. No per-app integration.
Fast on CPU
Parakeet engine runs 5–10× faster than Whisper on CPU. No GPU required. Long dictations stay snappy on Windows laptops.
99 languages + translation
Whisper engine handles 99 languages with the same hotkey. Translate-to-English mode turns any source language into English text on the fly.
100% local option
Whisper or Parakeet run entirely on-device. No audio leaves your machine. No account, no internet required for transcription.
AI rewriting
Optional Pro tier: GPT-5-mini cleans up filler, formats bullet lists, switches tone, drafts emails. Local Ollama is also supported.
Web search by voice
Ask a question out loud, get a researched answer typed into your document with sources. No browser switching. Pro tier.
Whisper vs Windows Voice Typing (Win+H)
Windows has shipped built-in dictation since Vista. Microsoft's Voice Typing (Win+H) is the current default. Here is how Whisper compares.
| Feature | Whisper by Remskill | Windows Voice Typing |
|---|---|---|
| Languages | 99 languages, including translate-to-English | Limited set; quality varies by language |
| Works in every app | Yes — any app, any field, hotkey global | Yes for most text fields |
| Offline mode | Fully offline with Whisper or Parakeet | Cloud-based; offline limited |
| AI rewriting | GPT or local Ollama: clean up, reformat, translate, change tone | None |
| Web search by voice | Ask questions, get researched answers with sources | None |
| Custom vocabulary | Hotwords, presets, custom dictionary per session | Limited |
| Price | Free local. Pro from $9.99/mo or $69 lifetime | Free with Windows |
Microsoft Voice Typing is fine for short notes and casual messages. Whisper exists for everything where built-in stops being enough — long-form writing, code dictation, non-English work, AI-assisted rewriting, and web research at the cursor.
99 languages, one hotkey
Whisper's engine speaks Polish, German, French, Spanish, Italian, Japanese, Korean, Mandarin, Arabic, Hindi, and 89 more. Code-switch mid-sentence and it still keeps up. The Parakeet engine adds English plus 24 European languages with much faster CPU performance — pick the one that matches your work.
Translate-to-English mode is built in: speak in any source language, get English text. Useful for journalists, researchers, and anyone collaborating across borders. All language switching happens locally — no cloud needed.
Your audio never leaves your machine
Local mode is the default. Audio is captured to a buffer, transcribed by Whisper or Parakeet running on your CPU, then discarded. No upload, no storage, no telemetry on transcript content. The desktop app is code-signed for Windows SmartScreen.
Cloud mode is opt-in and uses your own OpenAI API key (BYOK). Audio is sent only to OpenAI's API; Remskill never sees or stores it. This is documented in the Privacy Policy and verifiable in the open networking traffic from the app.
No telemetry on content
We don't log transcripts, audio, or what you type.
Open about networking
Local mode makes zero network calls during transcription.
BYOK for Cloud
Cloud mode uses your OpenAI key. Remskill takes no cut.
Code-signed installer
Standard Windows code-signing. No SmartScreen warnings.
How people use it on Windows
Speech-to-text on Windows is a category, not a single use case. Here is where Whisper users tell us it makes the biggest difference.
Email and replies
Dictate Outlook drafts, Gmail replies, Slack messages. The AI rewrite layer cleans up tone and grammar before paste.
Long-form writing
Word, Google Docs, Notion, Substack. Speak 200 words per minute, edit afterwards — most people write faster this way.
Customer support
Reply to tickets in Zendesk, Intercom, or Freshdesk faster. Custom presets keep tone consistent across the day.
Coding
Dictate prompts to Cursor, ChatGPT, or Copilot Chat without leaving the editor. Custom hotwords for variable and function names.
Meetings and notes
Capture thoughts between calls without typing. Useful when you're tired of meeting fatigue plus typing fatigue.
Translation
Speak in your native language, paste in English (or vice versa). Works for everything from emails to documentation.
Two transcription engines, your choice
We let you pick — neither is locked behind a paywall.
Parakeet (NVIDIA TDT, ~600 MB)
The fastest local option. 5–10× faster than Whisper on CPU. Best for English and 24 European languages where you need speed over flexibility.
- 5–10× faster than Whisper on CPU
- English + 24 European languages
- No GPU required
Whisper (OpenAI, 8 sizes)
The classic. 99 languages, translate-to-English, finer control with custom vocabulary, beam size, and hotwords. Slightly slower; pick when accuracy or multilingual work matters.
- 99 languages, translate-to-English
- Tunable: beam size, hotwords, vocabulary
- Best quality for long-form audio
Speech-to-text on Windows — questions
Is Whisper for Windows free?
Does it work offline?
Which Windows versions are supported?
How accurate is it compared to Windows Voice Typing?
Does it work in every app?
Can I dictate code?
What about privacy?
Try Whisper on Windows — free
Local pipeline is free for everyone. No card. No trial timer. Cloud features unlock when you want them.
Windows 10 and 11 (64-bit). 7-day Cloud trial included with Pro upgrade.