NSynth-MIDI-Renderer for massive MIDI dataset

This project is slightly modified from original work. It is designed to render large midifile datasets.

How is it different from the original work?

For a given MIDI sequence, it goes through a process of finding and assigning candidates to lock in a single preset from the beginning to the end of the sequence. If it fails to find one because the combination of instrument, pitch, and velocity is unique, it will randomly use a preset for every note.

New Version: https://github.com/spear011/NSynth-MIDI-Renderer-for-massive-MIDI-dataset
Original Project (last updated 2023/08/26): https://github.com/hmartelb/NSynth-MIDI-Renderer
When tested with a 130,000 MIDI dataset, only 95 MIDI sequences were randomized, with a probability of 0.0007%.

Quick start

Clone this repository to your system.

$ git clone https://github.com/hmartelb/NSynth-MIDI-Renderer.git

To start synthesizing audios, you need to download the NSynth dataset and MIDI dataset you want.

The NSynth Dataset, “A large-scale and high-quality dataset of annotated musical notes.” https://magenta.tensorflow.org/datasets/nsynth

How to use it

For the general use case, 3 parameters must be specified:

The path to the NSynth Dataset (dir path)
csv file that containing midi file id and instrument information
The output dir path

Midi Dataframe

The dataframe containing the information in the midi dataset should look like this.

id	instrument_str	split_data	...
midifile01	keyboard	train	...
midifile02	guitar	train	...
midifile03	guitar	val	...

Import the NoteSynthesizer Class into a Python script

from NoteSynthesizer import NoteSynthesizer

# Initialize variables here
# ...

# Create the NoteSynthesizer instance
synth = NoteSynthesizer(path_to_nsynth, path_to_midi_csv, output_dir, sr, velocities, preload)  

# you can use it with multiprocessing
import billiard as mp
from tqdm.auto import tqdm

cnt_cpu = mp.cpu_count() - 2
pool = mp.Pool(cnt_cpu)

output_dict = {'id': [], 'instrument': [], 'preset': [], 'source': []}
total = len(selected_file_list)

with tqdm(total=total) as pbar:
    for dict_object in tqdm(pool.imap_unordered(synth.render_sequence, selected_file_list)):
        for k, v in dict_object.items():
            output_dict[k].append(v)
        pbar.update()

Examples

The files in the /output folder looks like this.

output_dir/audio/train/midifile01_030_1.wav
output_dir/audio/train/midifile02_002_0.wav
output_dir/audio/val/midifile03_random_random.wav

The output dictionary looks like this

output_dict = {'id': ['midifile01', 'midifile02', 'midifile03'],
                'instrument': ['keyboard', 'guitar', 'guitar'] ,
                'preset': ['030', '002', 'random'],
                'source': [1, 0, 'random']}

The result of this dictionary can be used to track which presets and sources were selected for the MIDI file. Information about the presets of audio sources that were subsequently used can be matched with the nsynth dataset.

Citation

If you use this code in your research, please cite both projects as below:

@software{Martel_NSynth-MIDI-Renderer_2019,
    author = {Martel, Héctor},
    month = {8},
    title = {{NSynth-MIDI-Renderer: Sample based concatenative synthesizer for the NSynth dataset.}},
    url = {https://github.com/hmartelb/NSynth-MIDI-Renderer},
    version = {1.0.0},
    year = {2019}
}

@software{Han_NSynth-MIDI-Renderer-for-massive-MIDI-dataset_2023,
    author = {Han, ChangHeon},
    month = {8},
    title = {{NSynth-MIDI-Renderer-for-massive-MIDI-dataset : NSynth synthesizer for massive midi dataset}},
    url = {https://github.com/spear011/NSynth-MIDI-Renderer-for-massive-MIDI-dataset.git},
    version = {1.0.0},
    year = {2023}
}

Licence

MIT License
Copyright (c) 2019 Héctor Martel

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.