Overview / Usage

SongifyVox is an innovative voice-to-song conversion application that leverages cutting-edge AI and deep learning techniques to transform recorded voices into melodies or mesmerizing songs. The project aims to solve the challenge of creatively and effortlessly turning raw vocal inputs into professionally composed musical compositions.

Future Work: This AI application is about changing the recording to the desired melody and currently running on asian artists. The future work is to facilitate the Asian music industry by making it versatile.

Methodology / Approach

First of all, we trained models(AI artists) using versatile artists' songs and instrumental sounds using pytorch on Google Colab.

1. **Preprocessing: **The voice that is to be converted should be clean without instrumental sound.
2. Model Architecture: Researchers and developers choose a suitable model architecture based on the task's complexity and scale.
3. Frameworks and Libraries: AI language models are built using deep learning frameworks like PyTorch which offer the tools and infrastructure to design, train, and deploy large-scale models efficiently.
4. Training Process: Training involves optimizing the model's parameters using backpropagation and gradient descent algorithms. It's a computationally intensive process, usually requiring powerful GPUs and large voice computations.
5. Deployment: Once trained, the AI language model is made available for use through Fast API that accepts input and gives ai converted voice.
6. **Continuous Improvement: **I am continuously improving it so that it can give better output and if the input voice is not clean it should work on input voice.
7. **Application Interface: **I have provided the Web UI to this application from which it can make API call to Fast API.It will make a call with the input of
   1- Raw Voice
   2- Pitch Value
   3-Quality(Norma,Low,High)
   4-Artist
   After Conversion, it will give an output of converted in the form of mesmerizing song.

Technologies Used

Following technologies have been used in this project.

1. Google Colab for AI artist training.
2. Visual Studio Code for making APIs in Python(3.10)
3. Pytorch and torch audio, torch creep for voice conversion and training
4. Vultr Nvidia GPU with Intel CPU's for the deployment of the application.
5. HTML, CSS, Javascript, and Bootstrap for Web Interface
6. Docker for containerizing the application
7. Following Python libraries are installation
   torch
   torchaudio
   torchcrepe==0.0.15
   torchgen==0.0.1
   numpy==1.23.5

​

​

​

​

​

Repository

https://github.com/Rayan-Rasheed/SongifyVox-Voice-to-Song