Silence gives a voice to people who cannot use one. Point the phone at your face and mouth a sentence. The camera reads the shape of your mouth, turns it into text, and speaks it aloud in a natural voice.
It works by tracking your lips at 25 frames a second, cropping a 96×96 window around your mouth, and feeding that video to a visual speech recognition model. A language layer then repairs the output, because lip reading is ambiguous by nature; "p", "b" and "m" look identical on the mouth.
Built With
- accelerate
- auto-avsr
- avfoundation
- core-ml
- core-video
- espnet
- fastapi
- hugging-face-transformers
- ios
- mediapipe
- numpy
- openai-api
- opencv
- python
- pytorch
- sentencepiece
- speech
- swift
- swiftui
- uvicorn
- vision
- websockets
- whisper
- xcode
- xcodegen
Log in or sign up for Devpost to join the conversation.