Uttered

Many (incl. me) watch YouTube videos for learning languages. Lessons, talks, vlogs, series. But usually we just watch passively. If we don't understand something, we might skip it or rely on subtitles. I've also tried taking screenshots or notes, but hardly really used the language learned or reviewed the notes. After a couple days, nothing still lives in our memory.

At first, my idea was just to make a GPT that could help us practice speaking based on the context of any video link. After trying for a day, I realized it wasn't powerful enough to make a good product. Besides chat, there were many other relevant painpoints to be solved so that we could permanently master what we've learned.

Some people might not even know which videos are good content for learning foreign languages. So I curated some playlists based on my own language learning experience, and used AI to extract vocab/idioms/grammar with timestamps to help users understand the language easily, allow them to revisit the moments in video and save to notes in one click, and cached the analysis so that everyone could access for free.

In order for the AI tutor to have enough context about the videos, I used APIs to transcribe them (if no native transcripts available) and feed transcripts to live talk and text chat AI, and prompted AI to ask user questions based on the video content. I also had AI transcribe the live talk session and generate reports for users to learn from.

To aid memory and enable users to truly master the language, I added a feature of turning notes into visuals, as well as another feature of notes review reminders based on the forgetting curve.