Inspiration

Avoiding vulgar content is hard, especially when you enjoy the style of hip-hop songs. So we decided that we could combine our taste in music with lyricism which educates instead of rotting the brain with negative and worldly lyrics.

What it does

It receives a song as an input, extracts the instrumental from the mp3, transcribes the lyrics, uses the initial lyrics rhyming scheme, syllable count and pattern to redesign a new song using the initial structure, but incorporating the users theme - this is implemented by using the gemini api to generate the new lyrics, consequently this is passed to the elevenlabs api to generate the new voice, and adjusted to align to the beat using python modules.

How we built it

We started by creating a console application which would take the song and the theme as the inputs, and this soon evolved into our web app which we then polished to provide a clean User Experience.

Challenges we ran into

Due to us utilising many modules and apis, we ran into issues for outdated modules which could only run in earlier versions of python (3.8), this conflicted with the Gemini API which could only run in 3.9+. We also ran into issues when interacting with some APIs, for instance, we experimented with using SUNO AI to generate more realistic vocals, however due to the lack of an official API, we attempted to use third party APIs which should have worked but ended up leaving us with a 502 Server Error, which set us back a couple hours as we had to find a new way to get the same result. We also ran into credit issues due to us utilizing multiple AI Agents.

Accomplishments that we're proud of

We are very proud that we managed to integrate many APIs seamlessly to provide the user with an effective way to listen to their favourite song without sacrificing their listening preferences in regards to lyrics.

What we learned

We learned how to manage AI agents, interacting with many APIs and creating good looking, efficient webpages.

What's next for LyricForge

We plan to spend more time perfecting the vocal generation and time placement of each vocal line. Finally, we would love to release and host the platform live, and will do so after assessing user credit limits.

Built With

Share this project:

Updates