Player not loading? Watch on YouTube
OpenWhispr is presented as an open source dictation alternative to Wispr Flow and MacWhisper. The speaker demonstrates speech input in a text field and a terminal, then walks through model choices and app settings. He says it responds about as quickly as Wispr Flow and appears to remove spoken fillers, but the video does not include a controlled accuracy or speed comparison.
The local setup uses a downloaded OpenAI model shown as roughly 1.4 GB. The speaker also points out two NVIDIA model options and says local transcription does not require an API key. He enables GPU acceleration and reports that his local usage does not count against the cloud word allowance. These are observations from his setup, rather than hardware requirements or performance guarantees.
The settings tour covers microphone selection, tap-or-hold hotkeys, automatic pasting, and a dictionary for words that transcription tends to misspell. The app also accepts uploaded audio for transcription. The speaker briefly discusses meeting notes, Google Calendar integration, and a CLI, without demonstrating those workflows in detail.
Model settings include the app's cloud service, other cloud providers through API keys, and a self-hosted option. The speaker describes local processing as private because it keeps data on the computer. He also points out a floating transcription preview before cleanup. The walkthrough focuses on dictation and model selection; it does not establish minimum hardware specifications or test every feature.
