“Music is the literature of the heart; it commences where speech ends.”
― Alphonse de Lamartine, Author
In a blooming world, having the art of music with the emerging songs style, types and genres and thousands of songs and music pieces each with million different beats and the expansion of all music and songs datasets nowadays, splitting the music source into stems, generating lyrics and classifying genre has become easy and able for all users.
Stems are simply units of any given audio signal as per the musical jargon; mixing these units, or as we now know, stems, produces complex and harmonized sounds.
Us human beings are superb in isolating these audio signals and process only the sounds we need and suppress those we don’t need to hear. Humans' minds are really magnificent and capable of doing astonishing stuff, yet, with the advancement of technology, it is possible to record or create sounds that we can hear later. Now, various sectors deal with manipulation and study of the sound signals that require stems, thus in that sense, we need algorithms that can separate audio signals effectively.
Since we can pick up little inconsistencies in the sound, the stem separation must be as clear and crisp as possible. Hence, we would harness the capabilities of deep learning, masking and regeneration to recreate individual stems.
Music tags is a set of descriptive keywords that carry high-level information about a music clip; those keywords could reveal information about emotion, genre and instrumentation.
Hence, tags could be used for music recommendation and discovery.
MusicTaggingDemo.mp4
You'll need to install Anaconda,
Node.js and Heroku CLI before
running the following script.
git clone https://github.com/ZeyadOsama/apollo.git
cd apollo
conda env create --file environment.yml
conda activate apollo
heroku local If you don't want to use Heroku CLI, you'll run the first script individually, then open two separate terminals and
run each of the following scripts in a terminal.
git clone https://github.com/ZeyadOsama/apollo.git
cd apollo
conda env create --file environment.ymlconda activate apollo
npm startconda activate apollo
cd server
flask runGo to samples for some audio files samples to try the whole project.
For all-time versions, please see the CHANGELOG file.
This project is licensed under the MIT License - see the LICENSE file for details