BetterWispr is a free, open-source voice dictation app for macOS, released under the Apache-2.0 license and positioned as an alternative to Wispr Flow. The workflow is simple: hold Option+Space (or use press-to-toggle with a shortcut of your choice), speak, release, and the transcribed text is typed into whichever app was in focus. If paste cannot safely reach the previous app, the text is copied to the clipboard instead, and recordings stop automatically after two minutes. Transcription runs on the Mac with Apple's on-device speech recognition, NVIDIA Parakeet v3 (listed for 25 European languages), Parakeet v2 (English) or Whisper large-v3 turbo; the Parakeet and Whisper models download once and then work offline. A vocabulary feature takes spelling hints and phrase replacements for names and jargon, and the app learns from corrections made in History or right after pasting. Spoken voice commands such as comma, new line and scratch that are supported. A meetings mode produces a live transcript of both sides of a call, with titles, summaries and action items generated on-device through Apple Intelligence or a local Ollama model. For users who want it, a bring-your-own-model option accepts Sarvam AI, Smallest AI or any OpenAI-compatible transcription endpoint; in that case audio is sent to the chosen endpoint, and the app does not fall back to a cloud provider on its own. Settings, vocabulary and history are stored locally in an unencrypted workspace.json file, and history can be turned off. Requirements are macOS 14 or later, with Apple Silicon recommended, and Accessibility and Microphone permissions. The project is written mostly in Swift and was created in early October 2026, with several 0.1.x releases published within days.