How LALAL.AI Assists Musicians in Launching an Online Karaoke Service

The fusion of technology and music has opened up new avenues for aspiring and seasoned musicians, karaoke enthusiasts, entrepreneurs, and even developers. LALAL.AI has also played a pivotal role in empowering individuals to realise their dreams and be creative. In this piece, we’ll share one of a myriad of such cases.

Back in 2022, four Fullstack Academy students created HARK, an online karaoke service empowered by new technologies, including LALAL.AI for creating backing tracks and the TensorFlow SPICE model for pitch detection.

For this article, we sat down with Chanelle Huang, a musician-turned-software-engineer and one of the co-founders of HARK to delve into the story of four musicians who built a karaoke app together as their capstone project.

🟡 How Did You Switch Your Career from Music to Software Engineering?

I lived in Los Angeles, doing my master's degree in classical music performance, where I was also teaching students. When COVID happened, I had to move back home to Hawaii. There, I was teaching a lot of my students remotely, just using Zoom, and learned a lot about the online environment, so I thought I could make a website for my students to schedule things and have practice notes.

Then I just started getting more and more into audio technology and web development. I started to learn coding on my own and found that I really liked it since I love working with problem-solving and puzzles. So from there, I decided to commit to coding.

🟡 How Did You Come Up With the Idea to Create a Karaoke Service?

Karaoke is something that I really enjoy doing. I love music and singing. My friends and I, who are non-musicians, like doing karaoke in our free time, too. So when I was thinking of music apps that I could make because my dream goal is to work in music tech, I was looking for different ideas related to music, and one of them was karaoke.

Recently, I graduated from Fullstack Academy, where I learnt a lot about full stack web development. There, I met a lot of my teammates with whom I worked on several projects. And since many of my boot camp peers are also musicians, I proposed the idea of making a karaoke website, which is something I've always wanted to do, and then stumbled upon LALAL.AI. We were very excited about it and explored a lot of AI technology, audio technology, and APIs.

I was the only classical musician and professional musician in my group, but speaking of other HARK co-founders, Joshua Reyes is a karaoke enthusiast, Hannah Abbasi is a former high school teacher, and Lisa Knox is a former dispatcher.

🟡 Tell Us a Bit More About HARK

HARK is an online karaoke service that allows users to select a song, listen to the original artist, sing along with the instrumental tracks and lyrics, and receive a score based on the accuracy of pitch compared to the original vocal track. We built the app as our capstone project.

HARK was actually a combination of our last names of the four people in the group. We were finding out ways to arrange our last name to come up with a funny name, and that’s how HARK was born. When we were looking it up, we thought it was cool that the meaning of this word was actually ‘to listen.’ So we thought that would be an appropriate name.

This was our final senior project, so we really wanted to make something that would show our passions but also showcase all the things that we had learned along the way about full stack web development. With this project, we got to explore a lot of technologies that aligned with our interest because before we were doing eCommerce websites, things that were very specifically designed or simply defined assignments that our coaches wanted us to do.

But HARK was our first project where we could really let loose and find things that we were interested in ourselves.

We don’t have any particular plans to develop it further, but I think if we could find more technologies that could score based on tone, for instance, then I think that would really up the level a lot. Now, it’s scored only based on pitch.

So far, the only users are just my friends.

🟡 Why Did You Choose LALAL.AI?

We found LALAL.AI on Google. I tried a few other similar services, and LALAL.AI was the most accurate and was the easiest to use. The quality of sound provided by other tools I had tried out was very poor. I also liked that the LALAL.AI’s interface was super easy, so during the work, there were no issues or challenges. TensorFlow.js was another technology that helped us with this project. It allowed us to compare pitches from microphone input to the original vocal track, created using LALAL.AI.

🟡 How Did LALAL.AI Help You with Your Project?

We used LALAL.AI to extract the vocal and instrumental tracks for the five songs we currently have on our app. We needed to separate the track so that we could have a user sing along to the instrumental one.

We needed the original vocal track as well so that we could get the pitch data from that because we were giving users a score based on their own pitch data versus the original track.

Then we just combined that with TensorFlow JS to pre-populate the data on the backend.

💡
Check our detailed guide on how to extract vocals from a track with LALAL.AI to later use it as a karaoke backing track.

🟡 What Advice Would You Give to Aspiring Entrepreneurs & Musicians Who Want to Join Music Tech?

I would advise aspiring entrepreneurs and musicians to not be afraid to explore technologies outside their comfort zone. I love encountering unique challenges while learning new technologies!


Follow LALAL.AI on Instagram, Facebook, Twitter, TikTok, Reddit, and YouTube for more useful information on all things music and AI.