AI Music Composer project with source code

Think of a world where music making is done by human musicians and computers that can think. The mix of computers and music has created something new and unique in the last few years. This article talks about the beautiful world of computer-made music and how computers learn from data and make music that we can enjoy.

The Birth of AI Music Composition

The journey of AI in music composition began with the goal of understanding and replicating the complexities of human creativity. Researchers and developers envisioned an AI system that could understand musical patterns, harmonies, and melodies, ultimately creating compositions that resonate with the emotional depth to those crafted by human composers.

The Role of Machine Learning (ML) in Music Composition

At the heart of AI music composition lies Machine Learning. ML algorithms, equipped with vast datasets of musical compositions, analyze patterns and structures to understand the sense of different genres and styles. This enables AI systems not only to replicate but also to innovate in music composition.

Real-World Applications of AI Music Composer

Classical Revival

AI composers have shown an impressive ability to imitate the styles of classical masters like Bach and Beethoven. The revival of classical compositions with a modern twist showcases the versatility of AI in maintaining and developing musical traditions.

Contemporary Creations

AI composers are not limited to the past. They excel in creating modern music, and adapting to evolving genres and trends. From electronic beats to indie pop, AI-generated compositions are breaking new ground and resonating with diverse audiences.

Challenges and Ethical Considerations

Striking the Right Balance

While AI music composition opens doors to endless possibilities, there’s a delicate balance to be maintained. Striking the right balance between innovation and keeping the authenticity of human creativity is a challenge that developers and musicians grapple with.

Ethical Dilemmas in AI-Generated Music

The question of authorship and ownership in AI-generated music raises ethical dilemmas. Who owns the rights to a piece composed by an AI? Exploring these ethical considerations becomes imperative as AI continues to play a more prominent role in the creative process.

AI Music Composer project code

To create an AI Music composer in python follow these steps

Step 1: GPU Assignment and Mounting Google Drive

				# ASSIGNING GPU
!nvidia-smi -L

# MOUNTING GDRIVE
from google.colab import drive
drive.mount('/content/gdrive')

# ASSIGNING GPU
!nvidia-smi -L

# MOUNTING GDRIVE
from google.colab import drive
drive.mount('/content/gdrive')

Step 2: Installing Jukebox

				# INSTALLING JUKEBOX
!pip install git+https://github.com/openai/jukebox.git

# INSTALLING JUKEBOX
!pip install git+https://github.com/openai/jukebox.git

Step 3: Importing Libraries and Setting Up Distributed Environment

				# IMPORTING LIBRARIES
import jukebox
import torch as t
import librosa
import os
from IPython.display import Audio
from jukebox.make_models import make_vqvae, make_prior, MODELS, make_model
from jukebox.hparams import Hyperparams, setup_hparams
from jukebox.sample import sample_single_window, _sample, \
                           sample_partial_window, upsample, \
                           load_prompts
from jukebox.utils.dist_utils import setup_dist_from_mpi
from jukebox.utils.torch_utils import empty_cache

# Setting up distributed environment
rank, local_rank, device = setup_dist_from_mpi()

# IMPORTING LIBRARIES
import jukebox
import torch as t
import librosa
import os
from IPython.display import Audio
from jukebox.make_models import make_vqvae, make_prior, MODELS, make_model
from jukebox.hparams import Hyperparams, setup_hparams
from jukebox.sample import sample_single_window, _sample, \
                           sample_partial_window, upsample, \
                           load_prompts
from jukebox.utils.dist_utils import setup_dist_from_mpi
from jukebox.utils.torch_utils import empty_cache

# Setting up distributed environment
rank, local_rank, device = setup_dist_from_mpi()

Step 4: Model Configuration

				# Model configuration
model = '5b_lyrics'
hps = Hyperparams()
# ... (setting various hyperparameters)
vqvae, *priors = MODELS[model]
vqvae = make_vqvae(setup_hparams(vqvae, dict(sample_length = 1048576)), device)
top_prior = make_prior(setup_hparams(priors[-1], dict()), vqvae, device)

# Model configuration
model = '5b_lyrics'
hps = Hyperparams()
# ... (setting various hyperparameters)
vqvae, *priors = MODELS[model]
vqvae = make_vqvae(setup_hparams(vqvae, dict(sample_length = 1048576)), device)
top_prior = make_prior(setup_hparams(priors[-1], dict()), vqvae, device)

Step 5: Selecting Mode and Setting Up Sample Hyperparameters

				# Selecting Mode
mode = 'primed'
codes_file=None
audio_file = '/content/gdrive/My Drive/primer.wav'
prompt_length_in_seconds=12

# Setting up sample hyperparameters
sample_hps = Hyperparams(dict(mode=mode, codes_file=codes_file, audio_file=audio_file, prompt_length_in_seconds=prompt_length_in_seconds))

# Selecting Mode
mode = 'primed'
codes_file=None
audio_file = '/content/gdrive/My Drive/primer.wav'
prompt_length_in_seconds=12

# Setting up sample hyperparameters
sample_hps = Hyperparams(dict(mode=mode, codes_file=codes_file, audio_file=audio_file, prompt_length_in_seconds=prompt_length_in_seconds))

Step 6: Metas and Labels Setup

				# Metas and Labels setup
metas = [dict(artist="Rick Astley", genre="Pop", total_length=hps.sample_length, offset=0, lyrics="YOUR SOUNG LYRICS HERE!")] * hps.n_samples
labels = [None, None, top_prior.labeller.get_batch_labels(metas, 'cuda')]

# Metas and Labels setup
metas = [dict(artist="Rick Astley", genre="Pop", total_length=hps.sample_length, offset=0, lyrics="YOUR SOUNG LYRICS HERE!")] * hps.n_samples
labels = [None, None, top_prior.labeller.get_batch_labels(metas, 'cuda')]

Step 7: Sampling Configuration

				# Sampling Configuration
sampling_temperature = .98
# ... (setting up various sampling parameters)

# Sampling Configuration
sampling_temperature = .98
# ... (setting up various sampling parameters)

Step 8: Sampling Process

				# Sampling Process
if sample_hps.mode == 'ancestral':
    # ... (ancestral sampling)
elif sample_hps.mode == 'upsample':
    # ... (upsample handling)
elif sample_hps.mode == 'primed':
    # ... (primed sampling)
else:
    raise ValueError(f'Unknown sample mode {sample_hps.mode}.')

# Sampling Process
if sample_hps.mode == 'ancestral':
    # ... (ancestral sampling)
elif sample_hps.mode == 'upsample':
    # ... (upsample handling)
elif sample_hps.mode == 'primed':
    # ... (primed sampling)
else:
    raise ValueError(f'Unknown sample mode {sample_hps.mode}.')

Step 9: Listening to Audio and Memory Management

				# Listening to Audio
Audio(f'{hps.name}/level_2/item_0.wav')

# Memory Management
if True:
    del top_prior
    empty_cache()
    top_prior=None
upsamplers = [make_prior(setup_hparams(prior, dict()), vqvae, 'cpu') for prior in priors[:-1]]
labels[:2] = [prior.labeller.get_batch_labels(metas, 'cuda') for prior in upsamplers]
zs = upsample(zs, labels, sampling_kwargs, [*upsamplers, top_prior], hps)
Audio(f'{hps.name}/level_0/item_0.wav')

# Listening to Audio
Audio(f'{hps.name}/level_2/item_0.wav')

# Memory Management
if True:
    del top_prior
    empty_cache()
    top_prior=None
upsamplers = [make_prior(setup_hparams(prior, dict()), vqvae, 'cpu') for prior in priors[:-1]]
labels[:2] = [prior.labeller.get_batch_labels(metas, 'cuda') for prior in upsamplers]
zs = upsample(zs, labels, sampling_kwargs, [*upsamplers, top_prior], hps)
Audio(f'{hps.name}/level_0/item_0.wav')

AI music composers use technology and imagination to make new music. They work with machines that can learn from music and make their music. This makes music more fun and different, making us think about what it means and who makes it. We are constantly finding new ways to make music with AI, and we know that the best music comes from both humans and machines working together.

For Further Exploration

Can AI truly replicate the emotional depth found in human-composed music?

Absolutely. The sophisticated algorithms used in AI music composition can analyze and replicate emotional nuances present in various musical genres, creating compositions that resonate with genuine emotion.

How do musicians collaborate with AI in the creative process?

Musicians collaborate with AI by providing input, prompts, or themes to guide the AI’s composition process. It serves as a creative partner, offering suggestions and generating musical elements based on the given input.

What challenges do developers face in perfecting AI music composition?

Developers grapple with finding the right balance between innovation and preserving human authenticity. Additionally, ethical dilemmas surrounding authorship and ownership of AI-generated music present ongoing challenges.