Local-first dictation on Windows in 2026: a 6-point checklist (with Handy, Whisperstream, SpeakType, Vlow)

Plenty of dictation apps now run speech recognition on your own PC. Before you pick one, check:

1. Where does the audio go? On-device by default, or only after you configure a cloud engine?
2. Do you have to pick and download models yourself? Open-source tools like Handy (Whisper / Parakeet V3) and SpeakType (SenseVoice, Parakeet, Whisper) give you choice; others ship a default.
3. Is there cleanup, and where does it run? Whisperstream offers optional local cleanup (off by default); some apps polish in the cloud.
4. Can you edit selected text by voice? Select a sentence, say "translate to English" or "make it more formal".
5. Mixed-language speech? If you switch between Chinese and English mid-sentence, test that specifically.
6. Licensing / account: free with no account, free tier + optional paid plan, or subscription.

Disclosure: I work on Vlow, so take the next paragraph with that in mind.

Where Vlow fits: speech is recognised on your computer; press a talk key (Right Ctrl on Windows, F2 on Mac), speak, and clean text is pasted at your cursor. Select text and press F3 to translate or rewrite it by voice. Chinese-first with mixed Chinese/English. Unlimited local dictation is free. After a one-time email sign-in, new users can try Pro for 7 days plus 800 standard Cloud polish uses; whichever limit you hit first, that part stops. F3 rewrite and translation work during the trial. When the trial ends, the app returns to the free plan and does not charge you automatically. Windows 10/11: search "Vlow" in the Microsoft Store. Mac 1.3 is available from the official site (Apple silicon). Details: https://v-low.com

If you want open source and don't mind configuring models, Handy, SpeakType or Whisperstream are solid starting points.

— Jarvis Dou