An Android app for short-form creators seeking lyric-free background music, including an AI-generation experiment and a brief public release before development stopped.
What I currently believe
A personal app is not carried by features alone; release, monetization, and maintenance determine whether it continues.
Evidence
I built an Android app around short-form, lyric-free background music, integrated Firestore, experimented with AI generation through Modal and the Hugging Face API, and released it publicly.
What changed my mind
I expected feature development and AI generation to sustain the project, but advertising errors and the difficulty of continuing maintenance eventually ended my interest in the app.
The original idea
MusicLib was meant to provide short-form background music without lyrics for creators making something like YouTube Shorts. I first began working on it during the Diwali break in 2023, alongside my JEE preparation. The early version reached Firestore integration and could generate a few basic music tracks.
Returning to the project
After changing branches, I restarted development in December 2025 and returned to MusicLib as a more sustained Android project. The work gradually expanded from organizing music to experimenting with custom generation.
The AI-generation experiment
I added an option to generate music using AI. The Android app used a Modal deployment to rent a GPU through its free basic tier. A Python script then called the Hugging Face API to generate music. This made the project a practical experiment in connecting a mobile product to a remote generation workflow.
Release and testing
Alpha testing began on 11 January 2026. From 17 January, alpha and beta testing ran simultaneously and continued until 27 February. Most of the final development happened between 24 February and 6 March 2026. The first production build was published on 2 March, and the last update was published on 31 March 2026.
Why development stopped
I tried to implement advertising, but ads repeatedly failed to display correctly. I was never able to integrate them successfully, and over time I lost interest in continuing the application. MusicLib is therefore a released but paused project, not an actively maintained product.
The project taught me that shipping a feature is only one part of carrying a product. Distribution, monetization, reliability, and the motivation to maintain the system determine whether an experiment continues beyond its first release.