KeyScribe uses on-device speech recognition. After a speech model is downloaded, plain dictation can run without cloud speech-to-text, telemetry, or an account.
KeyScribe is local-first, not network-free in every situation.
Plain Dictation uses local speech recognition and local text cleanup.
Speech model weights are downloaded on demand, cached locally, and reused offline.
AI rewrite modes can call your configured provider. Turn rewrite off for fully local text processing.
Apple Dictation is a useful baseline. KeyScribe is for people who want local speech plus workflow control.