Voxt is a voice input and writing assistance tool for Mac users, with its core positioning being "speak once, paste polished text in any Mac app." By default it transcribes locally, also supports connecting to remote models with the user's own key, and uses App Branch to adjust tone and style by app.
In terms of use, Voxt aims to replace manual typing, turning natural speech into clear, structured written text, suitable for email, documents, chat, code comments, meeting notes and many other writing scenarios.
Main capabilities include:
- Speech-to-text and cleanup: click the shortcut to speak, click again to finish, automatically remove filler words, correct self-corrections, fix punctuation and numbers, and paste at the cursor.
- Rewriting: after selecting text, press fn+Control, use natural language instructions (such as "shorten" or "more direct") to rewrite the selection, preview the result in a card and then inject it as a replacement.
- Translation: select text and press fn+Shift to translate, no need to speak; you can also speak directly and paste the translated result. Supports 100+ languages and is optimized for mixed Chinese-English speech.
- AI chat: when no text is selected, what you say is used as the prompt, the answer is displayed in a card, you can press Continue to keep the conversation going, and it supports injection or copying.
- Note capture: press right ⌘ to open a floating note panel, record to-dos by voice, automatically generate a short AI title, supports To do, In Progress, Done statuses and priority, and can sync to Obsidian or Reminders, or drag into any app.
- Per-app enhancement (App Branch): group by app or URL, configure separate transcription prompts for each group, for example keeping chat casual, adding a subject and signature to email, and recognizing real parameters in the terminal.
- Meeting transcription: press fn+Option to start a real-time card, distinguish microphone and system audio between "me" and "the other party," support real-time translation, subtitles or recording mode, and save meeting details after finishing, including a summary and to-dos.
- Recording analysis: drag audio or video files into Files, supports MP3, MP4, M4A, WAV, AAC, MOV and other formats, up to 12 hours, automatically transcribes, distinguishes speakers and generates a complete meeting analysis, and the task queue supports priority, cancellation and retry.
How to use: after downloading Voxt, trigger different functions via shortcuts (such as fn, fn+Control, fn+Shift, fn+Option, right ⌘), and paste the result at the cursor in any Mac app. Local transcriptionBy default it is free and unlimited; remote models require the user to provide their own API key.
Use cases: developers dictating commit messages, PR summaries and code comments; writers dictating long-form text and then rewriting for tone and clarity; capturing decisions and action items in real time during meetings; cross-language team communication and translation; quick input for everyday chat, email and documents.
Limitations and notes: Voxt is a Mac app, and official support information for other platforms is not provided; the cost of remote models depends on the user's own API key, and Voxt itself does not charge a subscription fee; local processing can be fully offline, but remote models require a network; details such as custom cleanup strength and device compatibility are not yet provided.
