Gradio demo for MT3: Multi-Task Multitrack Music Transcription. To use it, simply upload your audio file, or click one of the examples to load them. Read more at the links below.
Pulled and adjusted from: This Huggingface Repository
Source codes: MT3: Multi-Task Multitrack Music Transcription | Github Repo
Another Dockerfile which may be awsome: https://github.com/jsphweid/mt3-docker
To run this docker container, Ubuntu 20.04 baremetal localhost is recomended. You can certainly choose to run whether baremetal or on a virtual machine, whether localhost or remotely under the same LAN, or even on a remote cloud computer, but there might be some issues. the installation will be exact the same.
If you have not installed docker, please run
sudo apt update && sudo apt install -y curl git
sudo curl -fsSL https://get.docker.com | sudo bash -s dockerIf you have a GPU, please run
sudo sh nvidia-container-runtime-script.sh
sudo apt-get install nvidia-container-runtimegit clone https://github.com/hero-intelligent/MT3-Docker.git
sudo docker build -t mt3 .Then run
sudo docker run -d --gpus=all --name=mt3 -p 7860:7860 mt3 If you have CPU only or want to run on Windows docker, then run
sudo docker run -d --name=mt3 -p 7860:7860 mt3 Finally, visit localhost:7860 and enjoy! If you run it remotely, please change localhost to the remote ip instead.
The quality of transcription will be better if a pure instrumental piece is inputed. If you want to transcribe a song with vocal, it is highly recomended to remove the vocal part using the UVR5.
UVR5, or Ultimate Vocal Remover GUI v5.5.0, is an AI powered and self-contained vocal remover, which is supported on both Windows and GUI Linux.
Check out more details at the Official Website and Github Repo and follow the instructions to have the software installed.