Live Caption and Voice Typing

Live Caption and Voice Typing both rely on the same offline live transcription setup, but they serve different surfaces: one shows floating captions, while the other sends text into another app.

Best for

• Showing system audio as a floating subtitle window
• Dictating directly into chat apps, documents, or forms
• Users who already understand the main recording workflow and only need one side capability

Before you start

• Finish Getting Started, or configure a working Live Record Model in Settings > Model Settings.
• If you want to send text into another app, the feature still depends on the same offline live transcription stack.

Where Live Caption starts

1. Open Live Record inside the Create workspace, or switch to the dedicated Tools > Live Subtitles panel.
2. Find the Live Caption or System Audio Captions toggle on that page.
3. If you only want floating subtitles for system audio, you do not need to start recording first. Turning it on is enough.
4. If you later start Live Record, both can run in parallel.

What Settings > Subtitle Settings controls

• It controls the floating caption window behavior: startup behavior, always-on-top, window positioning and drag-to-dock, click-through, font size, width, color, and background transparency.
• It does not provide the start toggle. The real entry points stay on the Live Record page and Tools > Live Subtitles panel.
• If Live Caption is already on but the window still does not appear, continue to FAQ and Troubleshooting.

How Voice Typing starts

1. Open Settings > Voice Typing or switch to the Tools > Voice Typing panel.
2. Turn on Voice Typing.
3. Assign a global shortcut on that page.
4. Choose either Push to Talk (Hold) or Toggle (Press once).
5. Configure situational presets, project vocabulary linking, and prolonged silence auto-commit under the secondary tabs.
6. If it still is not ready, review the readiness and dependency status shown there.

Situational awareness and vocabulary linking

• Application Context Awareness: Voice Typing can detect active application focus and apply tailored formatting presets (e.g. programming symbols for code editors, conversational punctuation for messengers).
• Project Scope & Hotwords Linking: Terms and corrections defined in the Unified Dictionary (Settings > Vocabulary) can be targeted directly to Voice Typing or tied to active project contexts.
• VAD Auto-Commit: Prolonged silence detected by Silero VAD automatically commits and types pending buffers without requiring manual shortcut triggers.

Push to Talk versus Toggle

• Push to Talk (Hold) works better for short bursts because capture only runs while you hold the shortcut.
• Toggle (Press once) works better for longer dictation because one press starts and the next press stops.
• In both modes, Voice Typing still depends on a working Live Record Model, any required VAD model, an available input device, and background warm-up.
• If Voice Typing is not ready yet, the same settings page tells you whether the blocker is the shortcut, model, VAD, input device, or runtime warm-up.
• Settings > Input Device > Keep Microphone Active is off by default. Turn it on if you want lower voice typing latency by leaving the microphone warm while idle; leave it off if you prefer Sona to release the microphone whenever you are not recording, dictating, or previewing input.

When to choose which one

• If your goal is to see system audio as floating text instead of sending it somewhere else, start with Live Caption.
• If your goal is to put spoken text into another app's input field, start with Voice Typing.
• If one of the features is enabled but does not behave the way you expect, go straight to FAQ and Troubleshooting.