Artificial Intelligence in Music, Sound, Art and Design : 12th International Conference, EvoMUSART 2023, Held As Part of EvoStar 2023, Brno, Czech Republic, April 12–14, 2023, Proceedings

This book constitutes the refereed proceedings of the 12th European Conference on Artificial Intelligence in Music, Sound, Art and Design, EvoMUSART 2023, held as part of Evo* 2023, in April 2023, co-located with the Evo* 2023 events, EvoCOP, EvoApplications, and EuroGP. The 20 full papers and 7 short papers presented in this book were carefully reviewed and selected from 55 submissions. They cover a wide range of topics and application areas of artificial intelligence, including generative approaches to music and visual art, deep learning, and architecture. Preface Organization Contents Long Talks LooperGP: A Loopable Sequence Model for Live Coding Performance Using GuitarPro Tablature 1 Introduction 2 Related Work 2.1 Sequence Models for Music Generation 2.2 Controllable Music Generation 2.3 Live Coding 2.4 DadaGP Dataset 3 Methodology 3.1 Defining a Loop 3.2 Extracting Loops 3.3 Training 3.4 Controllable Inference 4 Results 4.1 Training Performance 4.2 Loop Performance 4.3 Subjective Evaluation 5 Discussion 6 Conclusion and Future Work References Chordal Embeddings Based on Topology of the Tonal Space 1 Introduction 2 Tonal Space: Tonnetz Revisited 2.1 Euler's Tonnetz 2.2 Tonal Graph 2.3 Tonal Space 3 Embeddings of Pitches and Chords 3.1 Metric on Pitches 3.2 Metric on Chords 3.3 Theoretical Estimates of Embedding Dimension 4 Experimental Setup and Results 4.1 Chord Estimation 4.2 Automatic Music Generation 4.3 Music Transformation 5 Conclusion References Music Generation with Multiple Ant Colonies Interacting on Multilayer Graphs 1 Introduction 2 Related Work 3 Methodology 3.1 Multilayer Model 3.2 Ants System Description 3.3 Algorithm Description 3.4 Design Choices 4 Results 4.1 Example 1: On the Role of Different Pheromone Types 4.2 Example 2: On the Use of Different Objective Functions 4.3 Example 3: On the Scalabilty of the System 5 Conclusion References Automatically Adding to Artistic Cultures 1 Introduction 2 Applications of Generative Deep Learning 2.1 Generative Deep Learning in the Sciences 2.2 Generative Deep Learning in the Arts 2.3 Drawbacks to Generative Deep Learning 3 A New Emphasis for Generative AI 3.1 Ethical Issues 3.2 Stakeholders 4 Neuro-Symbolic Approaches 5 Decision Foregrounding for Generative Music 6 Conclusions References Extending Generative Neo-Riemannian Theory for Event-Based Soundtrack Production 1 Introduction 2 Background 2.1 Related Work 3 Extended Generative NRT 3.1 The Distribution of Admissible CGNROs 4 The GENRT Music Generation System 4.1 Event-Reflecting Musical Episodes 4.2 Chord Sequence Generation 4.3 Bass, Melody and Percussion Generation 4.4 Dampening the Music Between Events 5 An Illustrative Example 6 Conclusions and Future Work References Is Beauty in the Age of the Beholder? 1 Introduction 2 Art Appreciation Approaches 3 Helping People Release Their Inner Art Critic 3.1 Generating the Sculptures 3.2 Making an Engaging Environment 4 Results 4.1 Process Validation 4.2 Symmetry Preference 4.3 Participant Behaviour 5 Discussion 6 Conclusion References Extending the Visual Arts Experience: Sonifying Paintings with AI 1 Introduction 2 Literature Review 3 Methodology 4 Results 5 Conclusion and Future Work References Application of Neural Architecture Search to Instrument Recognition in Polyphonic Audio 1 Introduction and Related Work 2 Methods 2.1 Network Search Space 2.2 Baseline Networks 2.3 Tuners 3 Design of Experiments 3.1 Dataset 3.2 Setup 4 Results: Classification Performance 5 Results: Best Architectures 6 Conclusions and Future Work References AI-rmonies of the Spheres 1 Introduction 2 Concepts of Interest 2.1 Brief History of Astronomical Harmonies 2.2 Sonification and Data-Driven Music 2.3 Stellar Spectra 3 Unsupervised Auditory Exploration of Stellar Catalogs 3.1 Autoencoders 3.2 Sequential Chordification of MILES Stellar Library 3.3 Pipeline and Technical Details for the Musification of 4D Stellar Spectra Latent Space 3.4 OBAFGKM Evaluation 3.5 Synthetic Stellar Spectra from Musical Chords 4 Unsupervised Music Composition System Based on Deep Learning 4.1 RNN, LSTM Networks and Attention Mechanism 4.2 Generative Composition Using Dual LSTM with Attention Networks 4.3 Pitch Class Set Theory Cross-Match 4.4 Conclusion and Prospective References SUNMASK: Mask Enhanced Control in Step Unrolled Denoising Autoencoders 1 Introduction 1.1 Autoregressive Models 1.2 Non-Autoregressive Models 1.3 Trade-Offs Between Autoregressive and Non-autoregressive Approaches 1.4 SUNMASK, A Non-autoregressive Sequence Model 2 Method 2.1 Model Training 2.2 Convolutional SUNMASK 2.3 Transformer SUNMASK 2.4 Inference Specific Settings 3 Related Work 4 Experiments 4.1 Musical Evaluation 4.2 Text Datasets 4.3 Music Control 4.4 Text Control 5 Conclusion References SketchSynth: Cross-Modal Control of Sound Synthesis 1 Introduction 2 Related Work 2.1 Sound-Shape Associations 2.2 Sketch Recognition 3 Methods and Material 3.1 Sketch Dataset 3.2 Sound Dataset 3.3 Sketch-to-sound Mapping Using Deep Learning 3.4 Participants 3.5 Apparatus 3.6 Procedure 4 Analysis and Results 4.1 Model Evaluation 4.2 Sound Ratings 4.3 Survey Responses 4.4 Evaluation of Semantic Sound-Sketches 5 Discussion 6 Conclusion References Towards the Evolution of Prompts with MetaPrompter 1 Introduction 2 Related Work 2.1 Prompt Engineering 2.2 Tools for Prompt Construction 2.3 Prompt Generation 3 Approach 3.1 Creating Meta Prompts to Represent Spaces of Prompts 3.2 Exploring Spaces of Prompts in an Interactive Fashion 3.3 Implementation 4 Experimentation 4.1 Scenario 1: Study with Users 4.2 Scenario 2: Variety and Convergence 5 Conclusions and Future Work References Is Writing Prompts Really Making Art? 1 Introduction 2 The Limitations of Linguistic Description 2.1 Levels of Control 2.2 Authorship 3 Data Implications 3.1 AI as Parasite 3.2 Parasitic Meaning 3.3 Data Laundering 4 Materiality and Embodiment 4.1 Embodiment in AI 4.2 Exposing Process 4.3 Material Agency 4.4 A New Medium? 5 Conclusion References Using GPT-3 to Achieve Semantically Relevant Data Sonificiation for an Art Installation 1 Introduction 2 Related Work 3 Implementation 3.1 Conceptual Development 3.2 Human Labeling of Sounds by Audio Elicitation 3.3 Generating Natural Language Interpretations of Data with GPT-3 3.4 Matching GPT-3's Data Interpretations to Sound Labels for Soundscape Generation 4 Results 4.1 Experiment 1: Preliminary Testing of Matching GPT-3 Data Interpretations to Sound Labels 4.2 Experiment 2: Extracting Emotional Qualities of the Data for Better Matching 5 Discussion 6 Conclusion References Using Autoencoders to Generate Skeleton-Based Typography 1 Introduction 2 Related Work 3 Approach 3.1 Data 3.2 Network Architecture 4 Results 4.1 Reconstruction of Skeletons 4.2 Latent Representation of Font Style 4.3 Exploring the Latent Space 4.4 Transforming Skeletons into Glyphs 5 Conclusion and Discussion References Visual Representation of the Internet Consumption in the European Union 1 Introduction 2 Related Work 3 Impacts of Consumption on the Environment 3.1 Data 3.2 Left and Right Panels 3.3 Central Panel 3.4 Interaction 4 User Testing 4.1 Interaction Test 4.2 Perception Test 5 Discussion 6 Conclusion References GTR-CTRL: Instrument and Genre Conditioning for Guitar-Focused Music Generation with Transformers 1 Introduction 2 Background 2.1 Symbolic Music Generation with Deep Learning 2.2 Controllable Symbolic Music Generation 3 Motivations 4 Conditioning Experiments 4.1 DadaGP Dataset 4.2 Model Description 4.3 Instrument Conditioning Experiment 4.4 Genre Conditioning Experiment 4.5 Examples 5 Evaluation Methods 5.1 Overall Pitch and Rhythmic Metrics 5.2 Instrumentation Metrics 5.3 Genre Metrics 6 Quantitative Analysis and Discussion 6.1 Pitch and Rhythm Metrics Results 6.2 Instrumentation Metrics Results 6.3 Genre Metrics Results 7 Subjective Analysis and Discussion 8 Conclusion and Future Work References Artistic Curve Steganography Carried by Musical Audio 1 Introduction/Background 1.1 Prior Work in Audio Steganography 1.2 Traveling Salesperson Art 1.3 Hamiltonian Cycles on Watertight Triangle Meshes 2 Curve Embedding in Audio 2.1 Formulation of Least Squares Problem 2.2 Shifting, Scaling and Re-parameterizing Targets for Better Fits 2.3 Storing Component Scales in Phase 2.4 Recovering Frame Alignments 3 Experiments 4 Discussion/Supplementary Material References LyricJam Sonic: A Generative System for Real-Time Composition and Musical Improvisation 1 Introduction 2 Related Work 3 Background 3.1 Variational Autoencoder (VAE) 3.2 Generative Adversarial Network (GAN) 4 Methodology 4.1 Training Spec-VAE 4.2 Training Text-CVAE 4.3 Training the GAN 4.4 Retrieval Module 4.5 Inference 5 Data and Evaluation 5.1 Data 5.2 Evaluation 6 Listening Tests 7 Conclusions References Searching for Human Bias Against AI-Composed Music 1 Introduction 2 Background 2.1 Bias Against Computer Composed Artwork 2.2 Bias and Musical Metacreativity 2.3 Influence of Context and Expectation When Searching for Bias Against AI-composed Music 3 Methods 3.1 Participants 3.2 Musical Excerpts 3.3 Procedure 4 Results 5 Discussion 5.1 Future Work 6 Conclusion References Short Talks Fabric Sketch Augmentation & Styling via Deep Learning & Image Synthesis 1 Introduction 2 Related Work 3 Dataset 3.1 Data Collection 3.2 Data Processing 4 Methodology 4.1 Sketch Augmentation 4.2 Sketch to Photo-realistic Image 4.3 Tiling 5 Experiments and Results 5.1 Discussion on Results 6 Conclusion 7 Future Work References Transposition of Simple Waveforms from Raw Audio with Deep Learning 1 Introduction 2 Related Work 2.1 Time-Scale Approaches 2.2 Phase Vocoder 2.3 Pitch Correction 3 Dataset 3.1 Waveform Generation 3.2 Input and Output 3.3 Train and Test Split 4 Experimental Design 4.1 Model 4.2 Evaluation 5 Results 5.1 Error in the Time Domain 5.2 Error in the Frequency Domain 5.3 Signal to Noise 6 Discussion 6.1 Limitations and Future Work 6.2 Novelty and Applications 6.3 Conclusion References AI-Aided Ceramic Sculptures: Bridging Deep Learning with Materiality 1 Introduction 1.1 Reference Works 2 Case Study: Psychedelic Forms - Text-to-Ceramics 2.1 Methodology 2.2 Translation Processes 3 Discussion on Agency and Creativity Within AI-Aided Processes 4 Conclusions References OSC-Qasm: Interfacing Music Software with Quantum Computing 1 Context 1.1 OSC Protocol 2 OSC-Qasm Server 2.1 GUI 2.2 CLI 2.3 Network Distribution 3 OSC-Qasm Clients 3.1 Max 3.2 Pure Data 3.3 SuperCollider 4 Closing Remarks References EvoDesigner: Aiding the Exploration of Innovative Graphic Design Solutions 1 Introduction 2 Related Work 3 Approach 3.1 Evolutionary Engine 4 Experimental Setup and Results 4.1 Memorising Posters 4.2 Evolving Dissimilar Posters 5 Conclusion References Improving Automatic Music Genre Classification Systems by Using Descriptive Statistical Features of Audio Signals 1 Introduction 2 Related Work 3 Methodology 3.1 Data Collection and Preparation 3.2 Feature Selection 3.3 Model Training 3.4 Model Evaluation 4 Results and Evaluation 5 Conclusion 6 Future Works References Musical Genre Recognition Based on Deep Descriptors of Harmony, Instrumentation, and Segments 1 Introduction 2 Related Work 3 Deep Semantic Features 3.1 Harmonic Properties 3.2 Instrument Predictions 3.3 Segment Statistics 4 Setup of Experiments 5 Discussion of Results 6 Conclusions References Author Index This book constitutes the refereed proceedings of the 12th European Conference on Artificial Intelligence in Music, Sound, Art and Design, EvoMUSART 2023, held as part of Evo_ 2023, in April 2023, co-located with the Evo_ 2023 events, EvoCOP, EvoApplications, and EuroGP. The 20 full papers and 7 short papers presented in this book were carefully reviewed and selected from 55 submissions. They cover a wide range of topics and application areas of artificial intelligence, including generative approaches to music and visual art, deep learning, and architecture