Skip to content

A small, open-source Mac dictation app that uses your ChatGPT sign-in. Speak into T3 Code, your editor, or any text field.

1.9 KB
Raw Blame

Privacy

GPT Voice has no developer-operated server, analytics, or account database. Transcription requires an internet connection and sends recorded audio to ChatGPT. It does not run an offline speech model.

Data Where it goes
Audio A private temporary folder on your Mac, then ChatGPT for transcription
ChatGPT session The existing Codex CLI configuration on your Mac
Transcript The selected text field, or your clipboard for manual paste
Transcript history ~/Library/Application Support/GPT Voice/history.json on your Mac
Preferences macOS preferences on your Mac

The app removes owned audio files after successful transcription, cancellation, and normal cleanup. A failed request can keep its audio for Retry until another recording starts or the app quits. A crash can leave temporary files until macOS removes them. Successful transcriptions are saved on this Mac with their date, time, and delivery result. When paste keystrokes are sent to a known app, history includes its name. The app cannot confirm whether that app accepted the text. Practice transcriptions, failed requests, and cancelled recordings are not saved. Use History to copy a transcript, delete one entry, or clear all entries. History is not uploaded by GPT Voice. The file is readable only by your macOS user account. Normal Mac backups can include this file. Audio is not kept in history. ChatGPT's handling of uploaded audio is subject to your account settings and OpenAI's policies.

Microphone access permits recording after you start dictation. Accessibility access permits automatic paste. You can decline it and paste with Command–V instead. The app rechecks permissions when you return from System Settings.

The website's interactive preview runs in your browser with fixed sample text. It does not record or upload audio. Its video is a staged demonstration, not an accuracy benchmark.