LipScribe Captions that listen and lip-read

LipScribe needs your camera and microphone to read your lips and hear you.

Start live captions Talk normally. Words appear as you speak; tap again to stop.
Room noise AUTO
Measured from your microphone
-- Waiting for sound
LipScribe adapts by itself: the louder the room, the more it reads your lips.

What each model heard

Ready
Reading your lips… Usually takes 2–3 seconds
  1. Clip sent to the Alienware
  2. Face found, mouth cropped
  3. Listening and lip-reading
  4. Writing both captions
Ears only twin_v5

Captions from sound alone appear here.

Listens to the sound only
Ears + lips ws_v5

Captions from sound and lips appear here.

Listens and lip-reads together

Who the model trusted, moment by moment

The last 3 seconds, scrolling as you speak
--sound
--lips
What the eyes saw 96×96 mouth crops the model used