Inspiration

During the hackathon's welcome ceremony, we were inspired by the sign language translator. We realized that during many events, a sign language translator is not available, so a real-time transcriber could be extremely helpful for both deaf audiences and the general population.

What it does

Hosts can create an event, and watchers can join by using a unique event ID. Panscriber streams and transcribes the video in real-time, translating to a language of each watcher's choice.

How we built it

We used Flask for the backend, and socket.io for streaming audio and subtitles between hosts and watchers. We used Materialize for styling.

Challenges we ran into

We had quite a bit of trouble getting audio formats to work with each other, since the MediaStream Recording API only outputs webm files, and Google Cloud Speech-to-Text does not work with webm. However, we resolved this by using the SpeechRecognition web API instead.

Accomplishments that we're proud of

Learning a lot about video streaming.

What we learned

We learned a lot about WebRTC and video streaming, as well as various Speech-to-Text APIs.

What's next for Panscribe

Share this project:

Updates