Vista is my Hackathon submission for Supabase Launch week 6. It was inspired by those animated subtitle while scrolling through Youtube Shorts, and I always wanted to play with ffmpeg, and this serves as a good opportunity!
This ends up a super challenging task:
I need to run ffmpeg for video encoding, but hosting a server will required alot of coding & maintainance, thus resorting to use ffmpeg-wasm, which could be use on modern browser that supports wasm.
Perform speech-to-text is not an easy task, to speed up MVP, I've utilized AssemblyAI API for the video transcription.
Because speech-to-text is an async task, I've combined Supabase Edge Function as webhook when the process is done, then use Supabase Realtime to populate the subtitle.