Team name: Team (this is our actual team name)

Inspiration

The idea of voice tracking karaoke actually came from thinking about streaming. Lots of people stream themselves singing on platforms like Twitch so we thought, what if the lyrics could catch and follow along live with the streaming as they sing? It won't be just a regular set track like Spotify’s, but rather something that actually listens to the streamer in real time and displays the lyrics based on their singing speed. The subtitles would be able to follow the streamer’s singing, even if their singing goes off-speed, which is common in a capella singing. The audience would then be able to fully enjoy the singing stream and follow along if they want. We wanted to build something that brings the streamer and their viewers closer together, interactive and fun.

What it does

Our app lets you pick a song and artist, then listens to you sing along. While you're singing, it compares your voice to the original lyrics and figures out what part of the song you're on. Then, it shows the current line on screen, updating as you go — so you’re always in sync with the music. Think of it like a super helpful lyric buddy that keeps up, even if you mess up a word or sentences. Whether you're practicing, performing, or just having fun, it makes singing feel smoother and more connected.

How we built it

Everything from backend to frontend is built with Python. It's intuitive user interface keeps things simple and straight to the point, hop on and start singing. While Spotify's API allows access to user data, it doesn't provide lyrics so we used Genius API. A real-time lyric matching algorithm compares the user's sung words to each line of the song's lyrics after identifying the currently playing track through Spotify's most recently played track and dynamically adjusts to the user's singing speed based on string similarity. AWS allows us to handle real-time requests like lyric syncing and Spotify API calls without downtime, and makes it easy to expand as usage grows.

Challenges we ran into

One of the hardest things we ran into was making sure the singer and the lyrics are parallel in real time. With limited processing power, and no credit card, our options were limited. So for the demo, we kept things simpler—we let users type in the lyrics manually to show how the matching part would work. It’s not fully live yet, but it proves the idea behind it works. With better hardware and more time, we know we could make the full version run smoothly.

What's next?

There is still much more to build for sound tracking. First, the real time voice tracking that can be parallel to the lyrics. Second, actually tracking the speed of singing then slow and fasten the background music while singing.