CRESENDO was born from my love for both music and AI. As a vocalist, I often found it difficult to access quality instrumental tracks that matched my pitch and pace—either they were too expensive, lacked customization, or simply didn’t exist for certain genres or moods. I wanted to build something that could empower creators like me—students, indie artists, and music enthusiasts—with an intelligent and affordable tool that generates tailor-made music on demand. The idea of combining artificial intelligence with musical creativity was too exciting to ignore, and that's how CRESENDO began.
What I learned: This project taught me how vast and beautiful the intersection of AI and music really is. I explored sound synthesis, tempo mapping, emotion detection in lyrics, and audio layering. I also learned to work with audio libraries, AI models for music generation, and how to make the user experience smooth and intuitive. Most importantly, I realized that building a tech product isn’t just about code—it’s about solving a real problem for real people.
How I built the project: CRESENDO was built in layers. I started by researching AI music models and experimenting with tools like Magenta and Riffusion. Then, I focused on creating a user-friendly interface where singers could input their lyrics or choose a mood or genre. The backend used AI to generate instrumentals based on the user’s input, and layered multiple tracks (drums, bass, melody, harmony) to create a full piece. I also added features for tempo adjustment, vocal pitch matching, and export options. Throughout, I used Python and integrated open-source music models along with libraries like PyDub and Librosa.
Challenges I faced: One of the biggest challenges was making the generated music sound human, not robotic. Early versions of CRESENDO produced rigid or dull music. It took time, experimentation, and learning to fine-tune the AI’s outputs. Another challenge was memory and processing time—generating music can be heavy, so I had to optimize how the models worked and even offload parts to the cloud. Designing the interface to be intuitive for non-tech users was also tricky but critical.
Log in or sign up for Devpost to join the conversation.