### Generate Music Samples with Pop Music Transformer (Python)

Source: https://github.com/yatingmusic/remi/blob/master/README.md

This Python script demonstrates how to use the PopMusicTransformer model to generate music. It shows examples of generating music from scratch and generating continuations from a given MIDI prompt. Dependencies include the REMI library and TensorFlow. The function takes parameters like checkpoint path, generation length, temperature, top-k sampling, and output path.

```python
from model import PopMusicTransformer
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0'

def main():
    # declare model
    model = PopMusicTransformer(
        checkpoint='REMI-tempo-checkpoint',
        is_training=False)
        
    # generate from scratch
    model.generate(
        n_target_bar=16,
        temperature=1.2,
        topk=5,
        output_path='./result/from_scratch.midi',
        prompt=None)
        
    # generate continuation
    model.generate(
        n_target_bar=16,
        temperature=1.2,
        topk=5,
        output_path='./result/continuation.midi',
        prompt='./data/evaluation/000.midi')
        
    # close model
    model.close()

if __name__ == '__main__':
    main()
```

--------------------------------

### Print Tempo Items

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Prints the first 10 'Item' objects representing tempo changes. Each item details the tempo's start tick and its corresponding BPM value.

```python
print(*tempo_items[:10], sep='\n')
```

--------------------------------

### Print Note Items

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Prints the 'Item' objects representing notes read from the MIDI file. Each item shows its name, start time, end time, velocity, and pitch.

```python
print(*note_items, sep='\n')
```

--------------------------------

### Read and Process MIDI Files

Source: https://context7.com/yatingmusic/remi/llms.txt

Loads MIDI files and extracts structured note and tempo information. The 'utils.read_items' function parses a MIDI file into lists of 'Note' and 'Tempo' items. Notes contain start, end, velocity, and pitch, while tempo items indicate BPM changes. Includes functionality to quantize notes to a specified grid.

```python
from utils import read_items, quantize_items, Item
import numpy as np

# Read MIDI file (assumes single instrument track)
note_items, tempo_items = read_items('song.midi')

# Each note_item contains:
# - name: 'Note'
# - start: start tick
# - end: end tick
# - velocity: MIDI velocity (0-127)
# - pitch: MIDI pitch (0-127)

print(f"Total notes: {len(note_items)}")
print(f"First note: {note_items[0]}")
# Output: Item(name=Note, start=0, end=480, velocity=64, pitch=60)

# Quantize notes to 120-tick grid (eighth notes at 480 ticks per beat)
quantized_notes = quantize_items(note_items, ticks=120)

# Tempo items contain tempo changes at each beat
for tempo_item in tempo_items[:5]:
    print(f"Tick {tempo_item.start}: {tempo_item.pitch} BPM")
```

--------------------------------

### Print Notes from MIDI Object

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Prints all notes from the first instrument in the parsed MIDI object. Each note is represented with its start time, end time, pitch, and velocity.

```python
print(*midi_obj.instruments[0].notes, sep='\n')
```

--------------------------------

### Quantize Note Items using utils

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Applies quantization to the 'Item' objects representing notes using the utils.quantize_items function. This adjusts note start and end times to align with a defined grid, often for rhythmic accuracy.

```python
note_items = utils.quantize_items(note_items)
```

--------------------------------

### Collect MIDI Files and Prepare Training Data (Python)

Source: https://context7.com/yatingmusic/remi/llms.txt

This snippet demonstrates how to collect all MIDI files from a specified directory using glob and then prepare this data for training a REMI model. It involves extracting events, converting them to word indices, segmenting into fixed-length sequences, and batching.

```python
import glob
import numpy as np

# Assuming 'model' is an initialized REMI model object
# model = ... 

# Collect all MIDI files from directory
midi_paths = glob.glob('./my_dataset/**/*.midi', recursive=True)
print(f"Found {len(midi_paths)} MIDI files")

# Prepare training data
# - Extracts events from each MIDI file
# - Converts events to word indices
# - Segments into sequences of length 512 (x_len)
# - Groups into batches with memory segments
training_data = model.prepare_data(midi_paths=midi_paths)

print(f"Created {len(training_data)} training segments")
# Each segment shape: [group_size, 2, x_len]
# group_size=5, 2 for (input, target), x_len=512

# Save prepared data for later use
np.save('training_data.npy', training_data)

# Load and use saved data
loaded_data = np.load('training_data.npy')
model.finetune(
    training_data=loaded_data,
    output_checkpoint_folder='my-model-checkpoint'
)

model.close()
```

--------------------------------

### Import Libraries

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Imports the necessary libraries, miditoolkit for MIDI processing and utils for custom utility functions.

```python
import miditoolkit
import utils
```

--------------------------------

### Prepare Custom Training Data for REMI Model

Source: https://context7.com/yatingmusic/remi/llms.txt

Converts a collection of MIDI files into training-ready data segments for model fine-tuning using the `PopMusicTransformer` in training mode. This involves iterating through multiple MIDI files to prepare them as input for the model's training process.

```python
from model import PopMusicTransformer
from glob import glob
import numpy as np

# Initialize model in training mode
model = PopMusicTransformer(
    checkpoint='REMI-tempo-checkpoint',
    is_training=True
)

```

--------------------------------

### Generate Music Continuation with PopMusicTransformer

Source: https://context7.com/yatingmusic/remi/llms.txt

Continues an existing MIDI file by generating additional bars based on the prompt's style and pattern using the PopMusicTransformer model. It utilizes a chord-aware checkpoint and requires TensorFlow 1.14. Takes parameters for the number of bars, temperature, top-k sampling, output path, and the prompt MIDI file.

```python
from model import PopMusicTransformer
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model
model = PopMusicTransformer(
    checkpoint='REMI-tempo-chord-checkpoint',  # Use chord-aware checkpoint
    is_training=False
)

# Generate continuation from existing MIDI file
model.generate(
    n_target_bar=16,
    temperature=1.2,
    topk=5,
    output_path='./result/continuation.midi',
    prompt='./data/evaluation/000.midi'  # Path to prompt MIDI file
)

model.close()
```

--------------------------------

### Generate Music Continuation

Source: https://context7.com/yatingmusic/remi/llms.txt

Continues an existing MIDI file by generating additional bars based on the prompt's style and pattern. This allows for extending existing musical ideas.

```APIDOC
## Generate Music Continuation

### Description
Continue an existing MIDI file by generating additional bars based on the prompt's style and pattern.

### Method
Not directly an HTTP API, but a Python function call.

### Endpoint
`PopMusicTransformer.generate()`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
from model import PopMusicTransformer
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model
model = PopMusicTransformer(
    checkpoint='REMI-tempo-chord-checkpoint',  # Use chord-aware checkpoint
    is_training=False
)

# Generate continuation from existing MIDI file
model.generate(
    n_target_bar=16,
    temperature=1.2,
    topk=5,
    output_path='./result/continuation.midi',
    prompt='./data/evaluation/000.midi'  # Path to prompt MIDI file
)

model.close()
```

### Response
#### Success Response (200)
Generates a MIDI file at the specified `output_path`, continuing the provided `prompt` MIDI.

#### Response Example
(A MIDI file named `continuation.midi` is created.)
```

--------------------------------

### Generate Music from Scratch

Source: https://context7.com/yatingmusic/remi/llms.txt

Generates a new piano music composition without any input prompt using the pre-trained PopMusicTransformer model. This is useful for creating entirely novel musical pieces.

```APIDOC
## Generate Music from Scratch

### Description
Generate new piano music composition without any input prompt using the pre-trained PopMusicTransformer model.

### Method
Not directly an HTTP API, but a Python function call.

### Endpoint
`PopMusicTransformer.generate()`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
from model import PopMusicTransformer
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model with pre-trained checkpoint
model = PopMusicTransformer(
    checkpoint='REMI-tempo-checkpoint',
    is_training=False
)

# Generate 16 bars of music from scratch
model.generate(
    n_target_bar=16,           # Number of bars to generate
    temperature=1.2,            # Controls randomness (higher = more random)
    topk=5,                     # Top-k sampling parameter
    output_path='./result/from_scratch.midi',
    prompt=None                 # None = generate from scratch
)

model.close()
```

### Response
#### Success Response (200)
Generates a MIDI file at the specified `output_path`.

#### Response Example
(A MIDI file named `from_scratch.midi` is created.)
```

--------------------------------

### Generate Music from Scratch using PopMusicTransformer

Source: https://context7.com/yatingmusic/remi/llms.txt

Generates a new piano music composition without an input prompt using the pre-trained PopMusicTransformer model. Requires TensorFlow 1.14 and specifies CUDA device. Outputs a MIDI file and takes parameters for number of bars, temperature, and top-k sampling.

```python
from model import PopMusicTransformer
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model with pre-trained checkpoint
model = PopMusicTransformer(
    checkpoint='REMI-tempo-checkpoint',
    is_training=False
)

# Generate 16 bars of music from scratch
model.generate(
    n_target_bar=16,           # Number of bars to generate
    temperature=1.2,            # Controls randomness (higher = more random)
    topk=5,                     # Top-k sampling parameter
    output_path='./result/from_scratch.midi',
    prompt=None                 # None = generate from scratch
)

model.close()
```

--------------------------------

### Fine-tune PopMusicTransformer on Custom MIDI Data

Source: https://context7.com/yatingmusic/remi/llms.txt

Adapts the pre-trained PopMusicTransformer model to a custom musical style by fine-tuning on user-provided MIDI files. It requires TensorFlow 1.14 and loads custom MIDI data using glob. The process involves preparing data, creating an output folder for checkpoints, and then initiating the fine-tuning process which trains for up to 200 epochs.

```python
from model import PopMusicTransformer
from glob import glob
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model in training mode
model = PopMusicTransformer(
    checkpoint='REMI-tempo-checkpoint',
    is_training=True
)

# Load your custom MIDI files
midi_paths = glob('./my_music_folder/*.midi')

# Prepare training data from MIDI files
training_data = model.prepare_data(midi_paths=midi_paths)

# Create output folder for checkpoints
output_checkpoint_folder = 'my-custom-model'
if not os.path.exists(output_checkpoint_folder):
    os.mkdir(output_checkpoint_folder)

# Fine-tune the model (trains for up to 200 epochs)
model.finetune(
    training_data=training_data,
    output_checkpoint_folder=output_checkpoint_folder
)

model.close()

# Note: After fine-tuning, rename your chosen checkpoint to "model"
# and copy dictionary.pkl from the original checkpoint folder
```

--------------------------------

### BibTeX Citation for Pop Music Transformer Paper

Source: https://github.com/yatingmusic/remi/blob/master/README.md

This is the BibTeX entry for the 'Pop Music Transformer: Beat-Based Modeling and Generation of Expressive Pop Piano Compositions' paper. It contains author information, title, publication year, publisher, and other relevant details for academic citation.

```bibtex
@inproceedings{10.1145/3394171.3413671,
  author = {Huang, Yu-Siang and Yang, Yi-Hsuan},
  title = {Pop Music Transformer: Beat-Based Modeling and Generation of Expressive Pop Piano Compositions},
  year = {2020},
  isbn = {9781450379885},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3394171.3413671},
  doi = {10.1145/3394171.3413671},
  pages = {1180–1188},
  numpages = {9},
  location = {Seattle, WA, USA},
  series = {MM '20}
}
```

--------------------------------

### Convert REMI Events to MIDI File

Source: https://context7.com/yatingmusic/remi/llms.txt

Generates playable MIDI files from REMI event sequences or model output. It supports creating new MIDI files from scratch or appending generated events to an existing MIDI file as a continuation. Requires the 'utils' module and a 'dictionary.pkl' file for event-to-word conversion.

```python
from utils import write_midi
import pickle

# Load dictionary for word-to-event conversion
checkpoint = 'REMI-tempo-checkpoint'
event2word, word2event = pickle.load(open(f'{checkpoint}/dictionary.pkl', 'rb'))

# Example: words from model generation
# words is a list of integer indices representing events
words = [15, 234, 567, 123, 890, ...]  # Model-generated word sequence

# Write MIDI from scratch
write_midi(
    words=words,
    word2event=word2event,
    output_path='output.midi',
    prompt_path=None  # None = create new MIDI
)

# Write MIDI continuation (appends to existing prompt)
write_midi(
    words=words,  # Only the newly generated words
    word2event=word2event,
    output_path='continuation_output.midi',
    prompt_path='prompt.midi'  # Append to this MIDI file
)
```

--------------------------------

### Extract Chord Progressions from MIDI using chord_recognition

Source: https://context7.com/yatingmusic/remi/llms.txt

Automatically detects and extracts chord progressions from MIDI note sequences using chord recognition algorithms. It initializes a MIDIChord model, reads MIDI files, quantizes note items to a grid, and then extracts chord progressions. The output is a list of tuples, each representing a time interval and its corresponding chord name.

```python
import chord_recognition
from utils import read_items, quantize_items

# Initialize chord recognition model
chord_detector = chord_recognition.MIDIChord()

# Read MIDI file and extract note items
note_items, tempo_items = read_items('input.midi')

# Quantize notes to align with grid
note_items = quantize_items(note_items, ticks=120)

# Extract chord progressions
# Returns list of [start_tick, end_tick, chord_name] tuples
chords = chord_detector.extract(notes=note_items)

# Example output format:
# [[0, 1920, 'C:maj'], [1920, 3840, 'G:maj/B'], [3840, 5760, 'Am:min']]
for start, end, chord in chords:
    print(f"Time {start}-{end}: {chord}")
```

--------------------------------

### Temperature Sampling for Music Generation

Source: https://context7.com/yatingmusic/remi/llms.txt

Controls the randomness and creativity of music generation using temperature-controlled sampling. The `PopMusicTransformer` model's `generate` method accepts a 'temperature' parameter to influence the output. Lower temperatures produce more predictable music, while higher temperatures lead to more diverse and experimental results.

```python
import numpy as np

def generate_with_different_temperatures():
    from model import PopMusicTransformer

    model = PopMusicTransformer(
        checkpoint='REMI-tempo-checkpoint',
        is_training=False
    )

    # Conservative generation (low temperature = more predictable)
    model.generate(
        n_target_bar=8,
        temperature=0.8,      # Low temperature
        topk=3,               # Limited choices
        output_path='./result/conservative.midi',
        prompt=None
    )

    # Balanced generation
    model.generate(
        n_target_bar=8,
        temperature=1.2,      # Medium temperature
        topk=5,
        output_path='./result/balanced.midi',
        prompt=None
    )

    # Creative generation (high temperature = more random)
    model.generate(
        n_target_bar=8,
        temperature=1.8,      # High temperature
        topk=10,              # More choices
        output_path='./result/creative.midi',
        prompt=None
    )

    model.close()

# Temperature controls the softmax distribution:
# - Lower (0.5-1.0): More conservative, follows training patterns closely
# - Medium (1.0-1.5): Balanced between predictability and creativity
# - Higher (1.5-2.0): More experimental, diverse outputs

generate_with_different_temperatures()
```

--------------------------------

### Print Tempo Changes from MIDI Object

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Prints the first 10 tempo changes found in the MIDI file. Each tempo change includes the BPM and the tick at which it occurs.

```python
print(*midi_obj.tempo_changes[:10], sep='\n')
```

--------------------------------

### Read MIDI to Item Objects using utils

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Reads a MIDI file and converts its notes and tempo changes into a list of 'Item' objects using the custom utils.read_items function. This provides a unified format for further processing.

```python
note_items, tempo_items = utils.read_items('./data/evaluation/000.midi')
```

--------------------------------

### Fine-tune Model on Custom Data

Source: https://context7.com/yatingmusic/remi/llms.txt

Adapts the pre-trained model to your own musical style by fine-tuning it on custom MIDI files. This allows for personalized music generation.

```APIDOC
## Fine-tune Model on Custom Data

### Description
Adapt the pre-trained model to your own musical style by fine-tuning on custom MIDI files.

### Method
Not directly an HTTP API, but a Python function call.

### Endpoint
`PopMusicTransformer.finetune()`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
from model import PopMusicTransformer
from glob import glob
import os

os.environ['CUDA_VISIBLE_DEVICES'] = '0'

# Initialize model in training mode
model = PopMusicTransformer(
    checkpoint='REMI-tempo-checkpoint',
    is_training=True
)

# Load your custom MIDI files
midi_paths = glob('./my_music_folder/*.midi')

# Prepare training data from MIDI files
training_data = model.prepare_data(midi_paths=midi_paths)

# Create output folder for checkpoints
output_checkpoint_folder = 'my-custom-model'
if not os.path.exists(output_checkpoint_folder):
    os.mkdir(output_checkpoint_folder)

# Fine-tune the model (trains for up to 200 epochs)
model.finetune(
    training_data=training_data,
    output_checkpoint_folder=output_checkpoint_folder
)

model.close()

# Note: After fine-tuning, rename your chosen checkpoint to "model"
# and copy dictionary.pkl from the original checkpoint folder
```

### Response
#### Success Response (200)
Saves fine-tuned model checkpoints to the specified `output_checkpoint_folder`.

#### Response Example
(Checkpoints are saved in the `./my-custom-model/` directory.)
```

--------------------------------

### Convert MIDI to REMI Events using utils

Source: https://context7.com/yatingmusic/remi/llms.txt

Transforms MIDI files into REMI event sequences, suitable for model input or analysis. This utility function reads MIDI, quantizes notes, extracts chords and tempo information, and then groups these elements before converting them into REMI event sequences.

```python
from utils import read_items, quantize_items, extract_chords, group_items, item2event

# Read MIDI file
note_items, tempo_items = read_items('input.midi')

# Quantize notes to grid
note_items = quantize_items(note_items)

# Calculate max time
max_time = note_items[-1].end

# Extract chords (optional, for chord-aware model)
chord_items = extract_chords(note_items)

# Combine all musical elements
items = chord_items + tempo_items + note_items

# Group items by bar
groups = group_items(items, max_time)

# Convert to REMI events
events = item2event(groups)
```

--------------------------------

### Chord Recognition and Analysis in MIDI

Source: https://context7.com/yatingmusic/remi/llms.txt

Analyzes MIDI files to detect chord progressions and their qualities, assigning a score to each detected chord. It utilizes the `chord_recognition.MIDIChord` class and requires notes in an 'Item' format. The output provides chord symbols (e.g., C:maj, G:dom/B) and their durations in beats.

```python
import chord_recognition
import miditoolkit
import numpy as np

# Initialize chord detector
detECTOR = chord_recognition.MIDIChord()

# Available chord qualities detected:
# - maj: Major chords (root, major third)
# - min: Minor chords (root, minor third)
# - dim: Diminished chords
# - aug: Augmented chords
# - dom: Dominant seventh chords

# Load MIDI file
midi_obj = miditoolkit.midi.parser.MidiFile('song.midi')
notes = midi_obj.instruments[0].notes

# Convert notes to items format
from utils import Item
note_items = [
    Item(name='Note', start=n.start, end=n.end,
         velocity=n.velocity, pitch=n.pitch)
    for n in notes
]

# Extract chords with scoring
chords = detector.extract(notes=note_items)

# Process detected chords
for start_tick, end_tick, chord_symbol in chords:
    duration_beats = (end_tick - start_tick) / 480
    print(f"{chord_symbol}: {duration_beats:.1f} beats")

# Example output:
# C:maj: 4.0 beats
# F:maj: 2.0 beats
# G:dom/B: 2.0 beats  (G dominant 7th with B bass)
# Am:min: 4.0 beats
```

--------------------------------

### Group Items by Time using Python

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Combines chord, tempo, and note items into a single list and groups them based on the maximum time. The `group_items` utility is used for this purpose. The grouped items are then printed to the console, with each group separated by a blank line.

```python
items = chord_items + tempo_items + note_items
max_time = note_items[-1].end
groups = utils.group_items(items, max_time)
```

```python
for g in groups:
    print(*g, sep='\n')
    print()
```

--------------------------------

### Extract Chord Progressions from MIDI

Source: https://context7.com/yatingmusic/remi/llms.txt

Automatically detects and extracts chord progressions from MIDI note sequences using chord recognition algorithms. This is useful for music analysis and understanding harmonic structure.

```APIDOC
## Extract Chord Progressions from MIDI

### Description
Automatically detect and extract chord progressions from MIDI note sequences using chord recognition algorithms.

### Method
Not directly an HTTP API, but Python function calls.

### Endpoint
`chord_recognition.MIDIChord().extract()`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
import chord_recognition
from utils import read_items, quantize_items

# Initialize chord recognition model
chord_detector = chord_recognition.MIDIChord()

# Read MIDI file and extract note items
note_items, tempo_items = read_items('input.midi')

# Quantize notes to align with grid
note_items = quantize_items(note_items, ticks=120)

# Extract chord progressions
# Returns list of [start_tick, end_tick, chord_name] tuples
chords = chord_detector.extract(notes=note_items)

# Example output format:
# [[0, 1920, 'C:maj'], [1920, 3840, 'G:maj/B'], [3840, 5760, 'Am:min']]
for start, end, chord in chords:
    print(f"Time {start}-{end}: {chord}")
```

### Response
#### Success Response (200)
Returns a list of tuples, where each tuple represents a chord progression with start tick, end tick, and chord name.

#### Response Example
```
Time 0-1920: C:maj
Time 1920-3840: G:maj/B
Time 3840-5760: Am:min
```
```

--------------------------------

### Parse MIDI File with miditoolkit

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Parses a MIDI file into a MidiFile object using the miditoolkit library. This object contains information about instruments and notes within the MIDI file.

```python
midi_obj = miditoolkit.midi.parser.MidiFile('./data/evaluation/000.midi')
```

--------------------------------

### Convert Items to Events using Python

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Transforms grouped items (chords, tempos, notes) into a structured event format. The `item2event` utility function handles this conversion. The first 30 events of the resulting list are then printed to the console, each on a new line.

```python
events = utils.item2event(groups)
```

```python
print(*events[:30], sep='\n')
```

--------------------------------

### Convert MIDI to REMI Events

Source: https://context7.com/yatingmusic/remi/llms.txt

Transforms MIDI files into REMI event sequences for model input or analysis. This utility is essential for preparing data for the REMI model.

```APIDOC
## Convert MIDI to REMI Events

### Description
Transform MIDI files into REMI event sequences for model input or analysis.

### Method
Not directly an HTTP API, but Python function calls.

### Endpoint
`utils.item2event()`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
from utils import read_items, quantize_items, extract_chords, group_items, item2event

# Read MIDI file
note_items, tempo_items = read_items('input.midi')

# Quantize notes to grid
note_items = quantize_items(note_items)

# Calculate max time
max_time = note_items[-1].end

# Extract chords (optional, for chord-aware model)
chord_items = extract_chords(note_items)

# Combine all musical elements
items = chord_items + tempo_items + note_items

# Group items by bar
groups = group_items(items, max_time)

# Convert to REMI events
events = item2event(groups)
```

### Response
#### Success Response (200)
Returns a list of REMI events, representing the musical content of the MIDI file in a format suitable for the REMI model.

#### Response Example
(A list of REMI event objects/dictionaries.)
```

--------------------------------

### Extract Chords using Python

Source: https://github.com/yatingmusic/remi/blob/master/midi2remi.ipynb

Extracts chord information from a list of note items. It utilizes the `extract_chords` utility function. The output is a list of chord items, which are then printed to the console, formatted with newlines.

```python
chord_items = utils.extract_chords(note_items)
```

```python
print(*chord_items, sep='\n')
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.