We are both multi-lingual but often face challenges when trying to watch videos from other countries that have not been manually translated. We instead wanted to automate this process so language is never a barrier to information.
What it does
Freedom of speech takes a video as an input and outputs a video in your native language. Special care was taken to accommodate for blank spaces in the video. We also normalise our voices to make them more natural sounding.
How we built it
We built this using ffmpeg, python, librosa, SciPy and NumPy.
Challenges we ran into
Normalising the voice to make it sound human. Stitching blank sounds into the output video to compensate for background music and other sound effects that cannot be translated.
Accomplishments that we're proud of
Being able to translate any video almost instantly (under 1 min of processing required)
What we learned
We learnt a lot of video and audio codecs and various ML algorithms used to interpret video is time and frequency domains
What's next for Freedom of Speech
Make it a cloud-hosted app so it is more accessible to people.