Voice input for AI prompts · Windows
Say it
messy.
Hold a hotkey, say what you need, let go — a clean prompt appears at your cursor. Filler words and mid-sentence corrections are fixed; every requirement and every “don’t” is kept.
Windows 11 · 60 free cloud minutes, no card · hold Ctrl + Win to talk (configurable)
Paste it
clean.
See it work
You ramble. Your AI gets a clear prompt.
set the max retries to five, no wait, make it three, and don't retry on four oh four errors
Set the max retries to three, and don't retry on 404 errors.
Real output from the default cleanup model, not a mock-up. It only tidies what you said: it never answers for you, never adds steps you didn’t ask for, and never presses Enter.
Built for
prompting
Not a note-taker. Raw keeps your exact words, Clear (default) removes filler and repetition, Structured turns a brain-dump into numbered tasks.
Fixes corrections
“Three, no wait, five” comes out as five. Filler and repeats go too.
Keeps every constraint
Requirements, conditions and every “don’t” survive cleanup unchanged.
Speaks developer
Mix languages freely. useEffect, React Query and pnpm spelled right; add your own glossary.
Pastes, never sends
Only into the window you started in. Your clipboard is restored.
Measured, not promised
Windows 11, streaming recognition: median of 54 test dictations when you pause briefly before letting go. The 36 cleanup cases (Chinese, English and mixed speech) check that requirements, negations and corrections survive; measured with the default cleanup model.
How it works
Three steps. Zero friction.
Say2Prompt lives in the system tray and works in whatever app you already use — a terminal running Claude Code, Cursor’s chat, ChatGPT in the browser.
Works with
- Claude Code
- Codex CLI
- Gemini CLI
- Cursor
- VS Code
- Windsurf
- ChatGPT
- Claude
- Gemini
- DeepSeek
- Windows Terminal
- any text box
Pricing
Free with your own keys.
Pro when you’d rather not manage keys at all.
$0
Forever free
- Unlimited dictation with your own API keys
- Fully offline voice typing
- All cleanup modes
- Personal glossary & history
- 60 minutes of free cloud dictation, one time
- No credit card required
No account needed for offline and BYOK use. A verified account is required for free hosted dictation.
$10/ month
or $96 / year — save 20%
- No API keys required
- 25 hours of cloud dictation per month
- AI cleanup included
- Up to 3 Windows PCs
- Unlimited offline and BYOK dictation
- 14-day first-purchase refund policy
Prices in USD; local taxes may apply at checkout. Cloud minutes are counted per dictation — up to 5 minutes each on Free and 10 on Pro, one at a time. Yearly Pro gets 1,500 minutes every month; unused minutes don’t carry over, and extra use is never charged automatically. See the Terms and Refund policy.
FAQ
Questions.
How is this different from ordinary voice typing?
Voice typing turns sound into words. Say2Prompt tidies what you say to an AI: it drops filler, resolves corrections, keeps every requirement and “don’t”, and can turn a brain-dump into numbered tasks. It pastes without sending, so you check before you press Enter.
Will it ever send my message for me?
No. Say2Prompt only pastes text. It never presses Enter, so nothing is sent or run until you choose to.
Which apps does it work with?
Any Windows app with a text box: terminals running Claude Code, Codex CLI or Gemini CLI, IDEs like Cursor and VS Code, AI chats in your browser, chat and office apps. The exception is apps running as administrator — Windows blocks pasting into them, so the text goes to your clipboard instead.
What do I need for the free plan?
Nothing to start: sign in once with your email and you get 60 minutes of cloud dictation, one time. After that, keep dictating for free with your own Soniox and Groq API keys or an on-device model — no account needed for those.
Can I change the hotkey?
Yes. Pick any key combination in Settings, and choose whether you hold it while talking, press once to start and again to stop, or both. Esc cancels.
What happens to my voice and text?
While you hold the hotkey, your audio is streamed to the speech provider (Soniox) and the recognized text is cleaned up by a language model provider (on cloud dictation Groq, or Alibaba Cloud Model Studio in mainland China when Groq is unavailable) — through our servers on cloud dictation, directly with your own keys. Nothing is used for training and Say2Prompt keeps neither the audio nor the text. With an on-device model and cleanup off, nothing leaves your PC. Read the Privacy policy.
What happens when my cloud minutes run out?
Cloud dictation pauses until your allowance renews (Pro: every month). You're never charged for extra use. You can switch to an on-device model or your own API keys at any time.
Does it work on Mac or Linux?
Windows only for now.
My browser or Windows warned me about the installer.
While a release is new and few people have downloaded it, Chrome can say the file “isn’t commonly downloaded” and Windows can show SmartScreen; this goes away as our code signing and download reputation are established. Only if the installer came from say2prompt.com (dl.say2prompt.com), choose “Keep” in the browser and “More info” → “Run anyway” in Windows.