AMT-CS4347

Automatic lyrics and melody trascriber for CS4347 This project is divided into two components, a back-end served by FastAPI and uvicorn, and a front-end built on React.

Front-end

Due to dependency issues, this can only be run on Python 3.7. The front-end can be run through the following steps.

1. From within this directory, run pip install -r requirements.txt
2. cd AMTYolo
3. Follow the instructions within the README file in the AMTYolo directory to set up MusicYolo
4. From within the AMTYolo directory, run uvicorn main:app --host 0.0.0.0 --port 80

Back-end

The back-end is built on React and runs on Node 14.15.3

1. From this directory, run the command cd ./Frontend/amt
2. Run npm install to install the project dependencies
3. Run npm start to start running the front-end open the web-app on your browser

Using the web-app

Voice recording

You can start recording your voice by clicking start, then end by clicking stop. Clicking the play button in the left column will let you hear your recording.

Upload a file

Clicking browse will allow you to choose any mp3 file to upload to the system.

Download from Youtube

Alternatively, you can also directly copy-paste a Youtube link here. Do note that this is the slowest option by far, since it requires downloading the audio before processing it. Clicking download automatically starts the vocal AMT process.

Passing to Back-end

After audio is collected from recording or uploading, the "Upload" button will appear at the bottom. Clicking it sends the most recently collected audio to the Back-end for AMT processing.

Additional Notes

Due to how the back-end handles small audio files (e.g. less than 1 second), such audio samples being sent to the back-end might break the application.