Auto Clean Song Version Conversion

Inspiration

As we were brainstorming, we wanted to create a program that would overall improve the daily lives of people. Our application facilitates the lives of adults who are around children by ensuring they don't have to worry about a song's profanity.

What it does

These are the steps the program follows to filter out curse words : Get song name and mp3 file from the user and get song lyrics Split mp3 into vocals and background music Audio map lyrics to vocals and get timestamps for each line Trim timestamps containing curse words Output clean song

How we built it

We built our program on python, using a variety of libraries, such as spleeter, aeneas, pydub, Genius.com, and Google Custom Search Engine API, Streamlit - Web application framework

Challenges we ran into

We ran into several challenges when developing our program. First, we had trouble getting SpeechRecognition to work on any of our computers, so we decided to not use that at all. We also had trouble getting Spleeter to work, mostly because it had a lot of dependencies. After getting that to work, we tried several frameworks to build the app GUI, but had trouble with all of them, until we decided on Streamlit.

Accomplishments that we're proud of

Overall, our program works very well. The song is censored correctly and the song flows relatively well. We also have a clean-looking GUI, which allows for facilitated use.

What we learned

We learned a lot about python libraries, not only the ones we used, but the ones we tried using. We didn't only learn how to apply them, but ultimately how they work to perform the job they do. We also learned about GUIs and how they work. Outside of hard skills, we learned about teamwork and collaboration. We were able to work together virtually, which was extremely hard.

What's next for Auto Clean Song Version Conversion

There are several advancements we want to make for our program. First, we want to develop it into a mobile application for facilitated use. We also want to integrate additional libraries to make the song flow better. Also, allow for youtube/online search of songs, instead of uploading an mp3 file. Beyond this by modifying the technology you could also make it live conversion where it could be used on microphones to not only block curse words but also potentially something from the audience that a program doesn’t want to be heard on a Livestream. Another example could be a kid-friendly YouTuber live streaming and a curse word accidentally slips then it being live could help.

Built With

aeneas
api
pydub
python
streamlit

Submitted to

HowdyHack 2021

Created by

I worked on the initial ideas for the program, along with helping mapping out the plan for how to go about the problem. Also attempted many different methods of getting the timestamps of the curse words. Along with helping build the presentation and project.

Noam Gariani
I worked on developing the GUI and finding libraries to work with the program and perform what we needed. I also helped on the presentation and edited the devpost information. Mostly worked alongside Joel Munoz.

Alexa Hernandez
I developed the algorithm and implemented the libraries need to filter out the profanity.

Shantanu Thorat
Provided ideas on python specific libraries that we would need for our project ideas and found ways to implement them. Also did this for the creation of our "app" and worked alongside Alexa Hernandez in this part of the project to troubleshoot the code and design.

Joel Munoz