Wait until each sentence is finished instead of showing it word by word. Applies to the screen and the audience readers. The patience slider above decides how long a pause counts as the end of a sentence: higher means the AI hears more before it finishes (better translations, no cut-off endings), lower means a snappier voice. Both apply from the next Start.

The AI voice is spoken on this device, for headphones or a PA zone. It comes from the same translation session and costs no extra plan hours. With both voice options off, a text-only engine is used: captions still stream word by word, but no voice is generated.

Streams the AI voice to the audience's phones; listeners press Listen on their reader page. Uses about 0.5 Mbit/s of upload while on, and costs no extra plan hours.

Speech that is already in a chosen language is normally skipped. With this on it is captioned as it is: Swedish speech gives Swedish text when Swedish is a chosen language. Costs the same as translation; with the voice playing here, the microphone pauses while the voice speaks. Applies from the next Start.

Only the enrolled speaker is translated; audience chatter, videos and the AI voice itself are ignored. Speak normally for a few seconds after Start to enroll. The voice profile lives in this tab only and is never stored.