A collection of MIR open-source tools, dedicated to efficient inference tasks.

Designed to be lightweight with minimal package dependencies, ensuring easy installation and compatibility with various environments. This reduces the complexity of managing dependencies and helps avoid potential conflicts with other packages.

Significantly accelerating the analysis and processing of large datasets. By leveraging multi-threading and multi-processing techniques, and using ONNX TRT/CUDA for multi-GPU acceleration, MUSION can efficiently handle multiple files simultaneously, reducing the overall processing time.

Making it easy to use across different tools and applications. Whether you prefer using the Python interface, command-line interface (CLI), MUSION provides a consistent and intuitive experience for all users.

beat beat tracking
Detect beat and downbeat in music

separate music source separation
Separate songs into "drums", "bass", "other", "vocals"

struct music structure analysis / music segmentation
Detect chorus part for pop songs

transcribe automatic music transcription
Transcribe piano, drums or vocal audio to BEAT-ALIGNED MIDI file.

mfl Mid-Focus Level
Determines the probability that a particular instrument dominates the song.

1. Using PyPI

python -m pip install musion

And you have to install madmom manullay due to some restrictions from the original package.

python -m pip install git+https://github.com/CPJKU/madmom

2. Download ONNX model files from https://zenodo.org/records/13906170, and put them in the corresponding folders.

All the tools use the same procedure and method, only different at tool name. Refer to the README.md of each tool for more details.
Here is a example for the struct tool.

from musion.struct import Struct

struct = Struct()

# Select one of the two ways for input
## 1. Process a single file, given its file path
struct_res = struct(audio_path='dir/audio.wav')
## audio_path could also be a directory that contains audio files, or a list of audio paths.
struct_res = struct(audio_path='dir/')
struct_res = struct(audio_path=['audio1.wav', 'audio2.wav'])

## 2. Process a single audio data, given its pcm
from musion import MusionPCM
pcm = MusionPCM(samples, sample_rate)
struct_res = struct(pcm=pcm)

""" Optional Parameters """
# Save the result to a file
from musion import SaveConfig
save_cfg = SaveConfig(dir_path='dir/to/save/in', keys=['struct'])
# Where the keys can be obtained by
struct.result_keys
# Save the result by passing save_cfg, to the file: dir/to/save/in/audio.strcut
struct(audio_path='dir/audio.wav', save_cfg=save_cfg)

# Enable parallel processing when input contains multiple files, just set a proper number for num_workers
struct(audio_path='dir/', num_workers=5)
# Overwrite the existing result
struct(audio_path='dir/', overwrite=True)

Almost the same parameters as above. Type '$ musion -h' for more details. Here's a comprehensive example.

$ musion separate test_wavs/ --save_dir results_dir/ --save_keys vocals.wav bass.wav --num_workers 5

If you would like to participate in the development of Musion you are more than welcome to do so. Don't hesitate to throw us a pull request and we'll do our best to examine it quickly.