Show HN: BetterWispr – Free, open-source dictation for Mac

BetterWispr is a free, open-source macOS dictation tool that runs speech models locally so transcriptions never leave the user’s Mac. It inserts cleaned text directly into any focused text field via a hold-to-talk shortcut (Option+Space by default) or a toggle key, supports macOS 14+ with Apple Silicon recommended, and offers multiple built-in models (Apple Speech, Parakeet TDT v3 and Japanese, Whisper Large v3 Turbo) plus optional API-compatible providers. The project is Apache 2.0 licensed, distributes model downloads on demand, and installs updates from GitHub Releases with signature verification via Sparkle. No account or cloud fallback is required; everything from raw audio to cleaned output is stored locally unless a user explicitly configures an external provider.

Functionally, BetterWispr offers adjustable cleanup levels (None, Light, Medium) to remove filler words, false starts, or perform clarity edits using on‑Mac models like Apple Intelligence or Ollama. It learns custom spellings and names via Vocabulary or History fixes, supports voice punctuation and commands (e.g., “new paragraph,” “scratch that”), and keeps a local history with both original and edited transcripts. A Notetaker mode labels microphone audio as Me and other app audio as Them and can generate summaries, decisions, and action items locally or via optional cloud endpoints. If direct insertion is blocked by an app, the transcript is copied to the clipboard so nothing is lost.