1 hour ago · Tech · hide · 0 comments

Handy (still from the project’s website demo). Handy is an open-source, cross-platform dictation app that transcribes speech locally using Whisper or Parakeet models and pastes the result into the active application. Its optional text post-processing can use either local or cloud models: Apple Intelligence on supported Macs, a custom OpenAI-compatible endpoint (including a local server), or providers such as OpenAI, Anthropic, OpenRouter and Groq. Speech recognition remains local, but a cloud post-processing provider receives the transcribed text. The built-in “Improve Transcriptions” prompt removes fillers, fixes spelling and punctuation, and converts spoken numbers and punctuation while preserving the original language and meaning. The result is cleaned-up dictation, much like Rambler for Android. Post-processing is optional and disabled by default, and its prompts can be customised. See also TypeWhisper and the speech recognition catalogue. The source code is available under the…

No comments yet. Log in to reply on the Fediverse. Comments will appear here.