BetterWispr is a free, open‑source dictation app for macOS 14 and later that performs speech‑to‑text entirely on the user’s Mac. By holding the Option + Space shortcut in any application and speaking, the app transcribes audio using locally installed models such as Apple’s on‑device speech recognition, NVIDIA’s Parakeet (covering English, Japanese and 25 European languages) or OpenAI’s Whisper Large v3 Turbo, then inserts the cleaned text where the user was typing【primary†L1-L20】【support-1†L1-L15】.
This image demonstrates the user interaction described in the first paragraph where holding Option+Space initiates transcription
Beyond basic transcription, BetterWispr offers a meeting‑notes mode that labels the user’s own speech as “Me” and other captured audio as “Them”; after recording, either Apple Intelligence on compatible Macs or a local Ollama model generates summaries, decisions and action items from the transcript【primary†L60-L70】【support-1†L35-L40】. The app also adapts to individual users: custom terms can be added in the Vocabulary pane or a misheard word corrected once in History or right after pasting, and BetterWispr remembers that spelling for future dictations【primary†L80-L90】【support-1†L25-L30】.
Because all processing stays on the device, users must manually download and manage language models, and transcription speed depends on Mac hardware—Apple Silicon machines are recommended for optimal inference【primary†L20-L25】【support-1†L15-L20】. Certain features are English‑specific, such as automatic filler‑word removal (which drops “ums” and false starts) and voice commands like saying “comma” or “new paragraph”; multilingual support relies entirely on the capabilities of the chosen speech‑recognition model【primary†L115-L125】. BetterWispr remains free and open source under the Apache 2.0 license, with updates delivered via GitHub Releases and verified by EdDSA signatures【primary†L130-L140】.
This image reinforces the article’s emphasis on privacy and local processing discussed in the section about on‑device transcription and hardware considerations