How to Record
There are three ways to record:
- Toggle shortcut — Press Cmd+Shift+R (default) to start recording. Press again to stop and transcribe.
- Hold-to-talk — Hold the Option key (configurable to Control or Function) to record. Release to stop and transcribe. Press any other key while holding to cancel.
- "Hey Tota" wake word — With the wake word enabled in Settings > Dictation > Behavior, just say "Hey Tota" (or "wake up" / "hey Mac") to start recording hands-free. Uses Apple's on-device speech recognition and requires the Speech Recognition permission.
While recording, Tota shows a live status indicator — on Macs with a notch it hugs the notch itself, with a waveform while you speak and a dot-wave animation while transcribing; on screens without one, a floating pill appears instead. A pulsing red indicator confirms audio is being detected. Recording survives switching apps with Cmd+Tab, so you can keep talking while you move between windows.
Model Tiers
Tota uses WhisperKit to run OpenAI Whisper models entirely on-device. The Base model ships with the app; additional models download from HuggingFace on first use and work fully offline afterward.
| Model | Size | Notes |
|---|---|---|
| Tiny (English) | ~39 MB | Fastest, lightest |
| Base (English) | Built-in | Ships with the app — no download needed |
| Small (English) | ~244 MB | Good balance of speed and accuracy |
| Medium (English) | ~769 MB | Higher accuracy |
| Large v3 | ~1.5 GB | Maximum accuracy, multilingual |
| Large v3 Turbo | ~1.5 GB | Large-model accuracy with faster inference |
To change models, go to Settings > Models, select a model, and click Load Model. Dictation keeps using the current model until you load the new one.
Languages and Translation
With a multilingual model loaded (Large v3 or Large v3 Turbo), Settings > Dictation > Transcription unlocks:
- Language — Choose from 13 languages (Spanish, French, German, Italian, Portuguese, Japanese, Chinese, Korean, Russian, Arabic, Hindi, Dutch, English) or Auto-detect.
- Translate mode — Switch the mode from Transcribe to Translate and Tota types the English translation of whatever language you speak — still entirely on-device.
- English spelling — Choose US or UK spelling (available on all models). UK keeps British spellings like colour, realise, and centre across transcriptions and Smart Formatting.
Auto-Stop on Silence
Optionally, Tota can end a recording for you after a pause. Turn on Auto-stop recording on silence in Settings > Dictation > Behavior and set the silence duration (0.5–10 seconds, default 2). This is what makes the wake word fully hands-free: speak, pause, and the transcription pastes on its own.
Silence Filtering
Short transcriptions that contain only hallucination artifacts (such as "silence," "thank you," "thanks for watching," or "the end") are automatically discarded.
