Nowadays, many interviews are transitioning to online, and it has become increasingly hard to keep eye contact with the camera naturally. Furthermore, many influencers are moving into larger spotlights with the rapidly-developing online network, meaning that these individuals will likely spend more time speaking in front of cameras. Whether you're an interviewee preparing for your next Zoom meeting or an influencer wanting to convey more emotion into your videos, Presently has your back.
What it does
Presently is a web app that utilizes Azure Video Indexer, Livepeer, OpenCV, and Mediapipe to bring you the best user experience and features such as sentimental emotion detection, speech-to-text analysis, and eye-contact trainer.
How we built it
We used Python with OpenCV and Mediapipe to build the eye-contact detector and Node.js for the sentimental and emotion detectors. The frontend was built using Express.js and React. We used Firebase for data storage and uploads.
Challenges we ran into
- We had a difficult time figuring out how to pass the processed video from the backend to the frontend
- Since we only had 24 hours, we had to decide which features to include in our MVP, resulting in a lot of features being cut
- We originally thought of using IBM's Speech-to-Text API, but our credit cards didn't work so we had to change to Azure Voice Indexer halfway
Accomplishments that we're proud of
- It was our first time learning to use Computer Vision, but we were able to get it working within 24 hours :)
- The front-end design was created using purely Tailwind CSS and turned out to be beautiful
- We were able to react quickly to research and switch to a similar API when the original API didn't work out ## What's next for Presently
- More features to come, including access to previous recordings, video sharing through the web app, WPM and filler-word analysis, fidget analysis, and more.