A macOS application for real-time speech transcription and translation. Trans² captures microphone audio, transcribes speech using Apple's Speech framework, and translates it into a target language using Apple's Translation framework — all in real-time.

• Real-time transcription — Live speech-to-text using Apple's Speech framework with progressive transcription
• Automatic translation — Instant translation of transcribed text via Apple's Translation framework
• Multi-target language support — Translate into up to 10 target languages simultaneously, with one-click language swap (⌘⇧S)

• Audio level visualization — Color-coded waveform display (green/orange/red)
• Always-on-top mode — Keep the window above other applications (⌘T)
• Subtitle mode — Movie-style subtitle overlay at the bottom of the screen, showing translation only (⌘D); lines auto-expire after 30 seconds

• Audio file transcription — Transcribe existing audio files (WAV, MP3, M4A, AIFF) via File → Transcribe Audio File (⌘O)
• Export — Save transcripts as text or subtitles (.srt / .vtt), and export audio recordings (.m4a)

brew install --cask kcrt/kcrt/transsquared

• macOS 26 or later
• Microphone access

• Use a good microphone — A dedicated external microphone or headset significantly improves accuracy over built-in laptop microphones.
• Keep a quiet environment — Background noise (air conditioning, music, conversation) degrades recognition quality. Use a directional microphone if possible.
• Speak clearly at a steady pace — Avoid speaking too quickly or mumbling. Slight pauses between sentences help the recognizer segment text correctly.
• Stay close to the microphone — Keep a consistent distance of 15–30 cm from the microphone. Too far reduces signal quality; too close may cause clipping.
• Check audio levels — Use the audio level indicator in Trans² to verify that your microphone is picking up speech at an appropriate level. Green is good; consistently red indicates the signal is too loud.

Trans² captures audio from a microphone input. To transcribe audio playing on your Mac (e.g., a video call, a lecture video), you have two options:

macOS has built-in Live Captions that can transcribe system audio directly. Go to System Settings → Accessibility → Live Captions and enable it. Note that Live Captions only displays the original language — it does not translate.

To route system audio into Trans² for both transcription and translation, use a virtual audio loopback driver such as BlackHole:

1. Install BlackHole (2ch is sufficient).
2. Open Audio MIDI Setup (in /Applications/Utilities).
3. Create a Multi-Output Device that includes both your speakers/headphones and BlackHole.
4. Set this Multi-Output Device as your system output.
5. In Trans², select BlackHole as the microphone input (via Transcription → Microphone menu).

This way, system audio is sent to both your speakers and Trans² simultaneously.

Trans² processes all speech and translation on your device. Your audio and text are not sent to external servers.

• Translation — Apple's Translation framework documentation states: "All translations using the TranslationSession class are processed on the user's device." Apple may collect API usage metrics (such as app bundle ID and language pair) but does not collect the original or translated content.
• Speech recognition — Trans² uses the on-device SpeechTranscriber / SpeechAnalyzer API with locally downloaded models. Audio data is processed entirely on your device.

Available languages depend on the models Apple provides for each framework. Language availability may vary by region and OS version.

• Speech recognition — See Apple Speech Framework documentation
• Translation — See Apple Translation Framework documentation

Trans² uses on-device models for both speech recognition and translation. Models are downloaded automatically when you select a language that is not yet installed — a cloud icon (☁️↓) indicates languages that need downloading, and a progress indicator appears during download.

• Translation models — Open System Settings → General → Language & Region → Translation Languages to view, download, or remove translation language packs.
• Speech recognition models — Managed automatically by the system. There is no user-facing UI to delete them. The system may reclaim storage for models that have not been used recently.

• Open System Settings → Privacy & Security → Microphone and ensure Trans² has permission.
• Check that the correct microphone is selected in the Transcription → Microphone menu.
• If using an external microphone, verify it is properly connected and recognized by the system (check System Settings → Sound → Input).

• Ensure System Settings → Privacy & Security → Speech Recognition permission is granted to Trans².
• The speech model for the selected language may still be downloading. Wait for the cloud icon (☁️↓) to disappear.

• The translation model may not be downloaded yet. Go to System Settings → General → Language & Region → Translation Languages and download the required language pair.
• Some language combinations may not be supported. Try a different language pair.

• See Tips for Better Recognition above.
• Try adding domain-specific words to Custom Vocabulary in Settings (⌘,). Note that custom vocabulary provides hints to the recognition engine and may not always be prioritized.
• Use Auto Replace rules in Settings to correct frequently misrecognized words.

Found a bug or have a suggestion? Please report it on GitHub Issues. You can also access this link from the app's Help menu.

Technical documentation (architecture, audio pipeline, translation pipeline, etc.) is available in the docs/ directory.

All rights reserved.