Appearance
Theme
Cache
Interface
Language used in the app menus. Translation target is set below.
Language
Translation results will be shown in this language.
Keyboard Shortcuts
Start / Stop
Space
Close panel
Esc
Microphone
Sensitivity
Medium
Adjust if quiet voices aren't detected
Audio Source
Source
Select a browser tab and check "Share audio". Whole-screen sharing may include notification sounds.
Server fetches the stream via yt-dlp + ffmpeg and feeds it into the pipeline. Works for YouTube, BBC live, podcasts, direct .mp3/.m4a links. Live streams play until you Stop.
Play the active session's captured audio on this device — for monitoring what another device (PC tab / server mic) is picking up. Disabled on the device that's sending audio to avoid feedback. ~250 ms behind live.
Interpreter
배경음 볼륨
10%
Volume of the source audio you're listening along to. Interpreter voice always overrides — source ducks to ~15% while it speaks.
TTS Speed
1.0 = native TTS speed. Applied immediately to the current clip.
Read each translation aloud. Adds ~0.5–6 s per segment depending on the TTS engine.
Engine to render translations. Switch anytime — first switch loads the model (a few seconds for VoxCPM2/Kokoro).
Voice style
Pick a voice style for the interpreter. Preview regenerates the canonical voice and applies it to all subsequent translations.
Preview
Live preview (experimental)
Preview model
Off by default — shows raw interim text that can be inaccurate. Applies on next session start.
Subtitle Quality
Filter mode
Low-confidence segments are dimmed. Very low ones are hidden.