echo

echo is a Mac dictation app that transcribes locally, then optionally polishes text through your own Claude or ChatGPT account. It removes filler words, adds punctuation, mixes 25 languages and pastes formatted text into the active app.

echo is a dictation app for Mac that turns speech into finished text. The user holds a key, talks, and on release the words appear already punctuated and cleaned of filler words and repetitions, pasted into whichever app had focus. The aim is to remove the editing pass that usually follows dictation, so whole paragraphs can be spoken instead of single sentences. The workflow has three stages. Speech is first transcribed locally on the Mac, taking a fraction of a second and working offline. If polishing is enabled, the transcript is then sent as text, never audio, to the user's own Claude or ChatGPT account, which tidies it in roughly 1.5 to 2.5 seconds and only when the text needs it. The result is pasted into the active app and formatted for it, so an email, a chat message and a terminal each get a different treatment. Settings let the user set a tone per app, such as casual in Slack, and define how specific names should be spelled. It detects and mixes 25 languages in one passage, which removes the need to switch language mid-session. The tool suits people who write a lot in Gmail, Slack, GitHub, VS Code, Notion or Linear and who already pay for Claude Pro, Max or ChatGPT Plus. Without either subscription it still works with polishing off, but the user must then say punctuation aloud. If Claude fails or the connection drops, the raw transcript is pasted and the menu explains why, so dictated text is not lost. Among alternatives, the built-in macOS dictation is free and adequate for occasional sentences, and the maker says so openly. It types fillers as spoken, needs punctuation dictated, and handles one language per session. echo targets heavier use. It requires Apple Silicon and macOS 14 or later, and downloads a speech model of about 600 MB on first launch. The privacy position is that audio stays on the machine and nothing passes through the maker's servers.