Despite recent advances in multimodal large language models (MLLMs), their ability to understand and interact with music remains limited. Music understanding requires grounded reasoning over symbolic scores and expressive performance audio, which general-purpose MLLMs often fail to handle due to insufficient perceptual grounding. We introduce MuseAgent, a music-centric multimodal agent that augments language models with structured symbolic representations derived from sheet music images and performance audio. By integrating optical music recognition and automatic music transcription modules, MuseAgent enables multi-step reasoning and interaction over fine-grained musical content. To systematically evaluate music understanding capabilities, we further propose MuseBench, a benchmark covering music theory reasoning, score interpretation, and performance-level analysis across text, image, and audio modalities. Experiments show that existing MLLMs perform poorly on these tasks, while MuseAgent achieves substantial improvements, highlighting the importance of structured multimodal grounding for interactive music understanding.
Existing AI Music composition tools are limited in generation duration, musical quality, and controllability. We introduce CoComposer, a multi-agent system that consists of five collaborating agents, each with a task based on the traditional music composition workflow. Using the AudioBox-Aesthetics system, we experimentally evaluate CoComposer on four compositional criteria. We test with three LLMs (GPT-4o, DeepSeek-V3-0324, Gemini-2.5-Flash), and find (1) that CoComposer outperforms existing multi-agent LLM-based systems in music quality, and (2) compared to a single-agent system, in production complexity. Compared to non- LLM MusicLM, CoComposer has better interpretability and editability, although MusicLM still produces better music.
Analyzed in the study is the symbolic meaning of the images of a cicada/grasshopper in the prose and poetry by Boris Poplavsky in the context of classical antiquity and of the Silver Age art. These insects are often understood as one image, which is the poet’s homage to the antique tradition; the image of a cicada/grasshopper is liminal, implying mediation between the two worlds and connection between the past and the present, which allows one to experience the condition of harmonic unity with the Universe and the feeling of natural primeval being. Both the cicada and the grasshopper are related to music, but the cicada is to a greater extent an Apollonian and sacral creature, a votary of the muses, whose singing is heard against the background of a bucolic summer landscape. Whereas the image of the cicada is closer to the pole of the Apollonian, the image of the grasshopper, in the case when they diff er, tends to the pole of the Dionysian, the exuberant creative element and the energy of inspiration. The grasshopper turns into a fantastic chimeric creature, the visible personification of other living (a grasshopper among gnomes and elves, the horse-automobile-grasshopper and the grasshopper-train associations). A special case is the symbolization of sounds produced by a cicada and a grasshopper as a force of its own: this chirr/rattle represents certain natural buzz, the thundering proto music of inspiration over which the poet has no power. While the holistic image of these insects in the spirit of antiquity was the sign of experiencing the integrity of being and their chirr was enchanting, here the exponentially intensified startling sounds contribute to the description of the Poet’s “muse” as having the appalling features of a giant insect, and his poetic oeuvre becomes not only the blessing but also a curse, in which B. Poplavsky shows his affinity to A. Blok.
Сучасний простір академічного вокального мистецтва позначається значним розмиванням національних відмінностей у міжнародному дискурсі. Водночас китайське академічне вокальне мистецтво значною мірою зберігає свою своєрідність і характеризується еклектичними запозиченнями західних музичних та композиторських традицій, яке триває вже протягом століття. Музикознавче дослідження внеску основоположників академічного вокального мистецтва новітнього часу в Китаї є одним із найбільш актуальних напрямів наукового пошуку. У статті розглядається постать викладача, диригента, композитора, науковця Хуан Цзи й акцентується увага на романсовій творчості видатного китайського митця новітнього часу, який заклав підвалини китайського романсу; коротко характеризується освіта, життєвий і творчий шлях Хуан Цзи та визначається місце романсів у ньому. Також стаття містить аналіз романсів «Квіти – не квіти» (слова поета епохи Тан Бо Цзюй), «Три бажання троянди» (слова Лун Мухун), «Весняні роздуми» (слова у стилі «нової поезії» Вей Ханьчжана). У статті також розглядається специфіка вокально-виконавської інтерпретації романсів Хуан Цзи, обумовлена як особливостями музичної тканини цих творів, так і мовою путунхуа, якою написані вірші для них. Робиться висновок про те, що синтез різних музичних традицій і національних культур у вокальній творчості Хуан Цзи обумовив створення унікальних та своєрідних романсів, які не лише протягом майже усього ХХ століття (за винятком 10-річного періоду «культурної революції») зберігали свою популярність у Китаї, але й здобули міжнародне визнання та новий виток популярності на зламі ХХ та ХХІ століть. Романси Хуан Цзи є різноманітними як за композиційною побудовою, так і за поетичною основою, тому композитор обґрунтовано вважається засновником китайського романсу.
In this paper, we explore the application of Large Language Models (LLMs) to the pre-training of music. While the prevalent use of MIDI in music modeling is well-established, our findings suggest that LLMs are inherently more compatible with ABC Notation, which aligns more closely with their design and strengths, thereby enhancing the model's performance in musical composition. To address the challenges associated with misaligned measures from different tracks during generation, we propose the development of a Synchronized Multi-Track ABC Notation (SMT-ABC Notation), which aims to preserve coherence across multiple musical tracks. Our contributions include a series of models capable of handling up to 8192 tokens, covering 90% of the symbolic music data in our training set. Furthermore, we explore the implications of the Symbolic Music Scaling Law (SMS Law) on model performance. The results indicate a promising direction for future research in music generation, offering extensive resources for community-led research through our open-source contributions.
Generating music that aligns with the visual content of a video has been a challenging task, as it requires a deep understanding of visual semantics and involves generating music whose melody, rhythm, and dynamics harmonize with the visual narratives. This paper presents MuVi, a novel framework that effectively addresses these challenges to enhance the cohesion and immersive experience of audio-visual content. MuVi analyzes video content through a specially designed visual adaptor to extract contextually and temporally relevant features. These features are used to generate music that not only matches the video's mood and theme but also its rhythm and pacing. We also introduce a contrastive music-visual pre-training scheme to ensure synchronization, based on the periodicity nature of music phrases. In addition, we demonstrate that our flow-matching-based music generator has in-context learning ability, allowing us to control the style and genre of the generated music. Experimental results show that MuVi demonstrates superior performance in both audio quality and temporal synchronization. The generated music video samples are available at https://muvi-v2m.github.io.
Music is inherently made up of complex structures, and representing them as graphs helps to capture multiple levels of relationships. While music generation has been explored using various deep generation techniques, research on graph-related music generation is sparse. Earlier graph-based music generation worked only on generating melodies, and recent works to generate polyphonic music do not account for longer-term structure. In this paper, we explore a multi-graph approach to represent both the rhythmic patterns and phrase structure of Chinese pop music. Consequently, we propose a two-step approach that aims to generate polyphonic music with coherent rhythm and long-term structure. We train two Variational Auto-Encoder networks - one on a MIDI dataset to generate 4-bar phrases, and another on song structure labels to generate full song structure. Our work shows that the models are able to learn most of the structural nuances in the training dataset, including chord and pitch frequency distributions, and phrase attributes.
The new edition of the tablature of Joannes de Lublin represents the state of the art in editorial practice, combining a critical and practical edition. It is discussed within the contexts of the contemporary corpus of tablatures in Poland and also Central European music, developing a series of current topoi: a new ordering and reconstructing of the material taking account of codicological research and functional purposes, the local reworking of the repertoire and compilatory practice of the scribes, unknown concordances or unica, counterpoint and intabulation techniques, syntactic-rhetorical properties, and historical didactics. The edition provides literal, but good, readable transcriptions.
In 1953 there was a projected collaboration between Dylan Thomas and Igor Stravinsky: an opera whose libretto was to be written by the poet and set to music by the composer. Unfortunately, due to Thomas’s death that same year, this plan never materialised. The aim of this article is to offer a detailed exploration of the context within which this project originated and what it entailed for the poet, especially in relation to the direction where Thomas’s work was taking him before his death. The opera with Stravinsky could have been the culmination of the poet’s work inasmuch as it brought together his love of sound, rhythm and music, the horror of the war, his fascination with language and the near-extinction of human life. Moreover, the artists were able to exchange a series of letters before Thomas’s death. The second contribution of our article is the publication of Stravinsky’s letters to Thomas, which have not been previously published. Whilst the narrative of their encounters and relationship has been shaped in certain ways by different writers, a close examination of the primary source materials—the accounts of friends, the content of the letters and the circumstances in which they were written help shed light on this highly promising, but uncompleted, project.
Bu araştırma, Türk Sanat Müziği ses eğitimi ve sözlü repertuvar eğitiminde türkü formunun kullanımına dair uzman görüşlerinin tespit edilmesi amacıyla yapılmış olup; “Ülkemizde Türk Sanat Müziği eğitimi veren Konservatuvar ve Güzel Sanatlar Fakültelerindeki Türk Sanat Müziği Ses Eğitimi bölümlerinde türkü formu kullanımı nasıldır?” probleminin çözümlenmesi bağlamında beş alt probleme yönelik veri toplanmıştır. Bu alt problemlere yanıt alabilmek için yarı yapılandırılmış bir görüşme formu hazırlanmıştır. Bu doğrultuda; ülkemizde mesleki müzik eğitimi veren yükseköğretim kurumlarından Konservatuvar ve Güzel Sanatlar Fakültelerindeki Türk Sanat Müziği Bölümlerine yönelik bir çalışma yürütülmüş olup Türk Sanat Müziği sözlü repertuvar eğitimi ve Türk Sanat Müziği Ses Eğitiminde türkü formunun kullanımına dair uzman görüşlerine başvurulmuştur. Araştırmanın çalışma gurubunu 2018-2019 eğitim öğretim yılında, rastlantısal yolla seçilen İstanbul, Afyon, Kocaeli ve Konya’daki Konservatuvar ve Güzel Sanatlar Fakülteleri Türk Sanat Müziği Ses Eğitimi bölümlerinde görev yapan, Türk Sanat Müziği Repertuvar Eğitimi ve Türk Sanat Müziği Ses Eğitimi derslerini yürüten 7 Öğretim Elemanı oluşturmaktadır. Bireysel ses eğitimi veya repertuvar dersini ilgili fakültelerde yürütmüş ve yürütmekte olan bu öğretim elemanlarına uygulanan yarı yapılandırılmış görüşme formu ile bilgiler elde edilmiştir. Literatür tarama ve ihtiyaç analizi doğrultusunda, betimsel tarama modeli ve nitel araştırma yöntemi kullanılarak ulaşılan veriler içerik analizine tabi tutularak sonuçlar kısmına ulaşılmıştır. Ulaşılan bu sonuçlar ışığında araştırmacı tarafından bazı öneriler sunulmuştur.
Igbo choral art music is a Nigerian genre that exhibits interesting musical hybridity as it combines a notated choral part with an un-notated orchestral part into a final musical artwork. The notated choral part is produced by the composer through a synthesis of Igbo traditional music and European classical music, while the orchestral part, based on Igbo traditional music practices, is produced by the choral director. For over five decades scholarly discussions on Igbo choral art music have focused on the notated choral part and how composers blend European classical music and Igbo traditional music principles. Although the choral part is significant, my more than two decades of participation in the genre as well as interviews with leading practitioners reveal that without the orchestral part, a composition is considered incomplete by both the practitioners and the audience. Such a critical aspect of the genre has received little to no academic attention. This article is an autoethnographic report on the orchestration principles of Igbo choral art music, its relationship with Igbo traditional music, and how such orchestration contributes to hybridity as a fundamental resource in the genre’s development.
Andreas Förster, Alarith Uhde, Mathias Komesker
et al.
LoopBoxes is an accessible digital musical instrument designed to create an intuitive access to loop based music making for children with special educational needs (SEN). This paper describes the evaluation of the instrument in the form of a pilot study during a music festival in Berlin, Germany, as well as a case study with children and music teachers in a SEN school setting. We created a modular system composed of three modules that afford single user as well as collaborative music making. The pilot study was evaluated using informal observation and questionnaires (n = 39), and indicated that the instrument affords music making for people with and without prior musical knowledge across all age groups and fosters collaborative musical processes. The case study was based on observation and a qualitative interview. It confirmed that the instrument meets the needs of the school settings and indicated how future versions could expand access to all students, especially those experiencing complex disabilities. In addition, out-of-the-box functionality seems to be crucial for the long-term implementation of the instrument in a school setting.
Vocal pitch is an important high-level feature in music audio processing. However, extracting vocal pitch in polyphonic music is more challenging due to the presence of accompaniment. To eliminate the influence of the accompaniment, most previous methods adopt music source separation models to obtain clean vocals from polyphonic music before predicting vocal pitches. As a result, the performance of vocal pitch estimation is affected by the music source separation models. To address this issue and directly extract vocal pitches from polyphonic music, we propose a robust model named RMVPE. This model can extract effective hidden features and accurately predict vocal pitches from polyphonic music. The experimental results demonstrate the superiority of RMVPE in terms of raw pitch accuracy (RPA) and raw chroma accuracy (RCA). Additionally, experiments conducted with different types of noise show that RMVPE is robust across all signal-to-noise ratio (SNR) levels. The code of RMVPE is available at https://github.com/Dream-High/RMVPE.
In the era of extensive intersection between art and Artificial Intelligence (AI), such as image generation and fiction co-creation, AI for music remains relatively nascent, particularly in music understanding. This is evident in the limited work on deep music representations, the scarcity of large-scale datasets, and the absence of a universal and community-driven benchmark. To address this issue, we introduce the Music Audio Representation Benchmark for universaL Evaluation, termed MARBLE. It aims to provide a benchmark for various Music Information Retrieval (MIR) tasks by defining a comprehensive taxonomy with four hierarchy levels, including acoustic, performance, score, and high-level description. We then establish a unified protocol based on 14 tasks on 8 public-available datasets, providing a fair and standard assessment of representations of all open-sourced pre-trained models developed on music recordings as baselines. Besides, MARBLE offers an easy-to-use, extendable, and reproducible suite for the community, with a clear statement on copyright issues on datasets. Results suggest recently proposed large-scale pre-trained musical language models perform the best in most tasks, with room for further improvement. The leaderboard and toolkit repository are published at https://marble-bm.shef.ac.uk to promote future music AI research.
Self-supervised learning (SSL) has shown promising results in various speech and natural language processing applications. However, its efficacy in music information retrieval (MIR) still remains largely unexplored. While previous SSL models pre-trained on music recordings may have been mostly closed-sourced, recent speech models such as wav2vec2.0 have shown promise in music modelling. Nevertheless, research exploring the effectiveness of applying speech SSL models to music recordings has been limited. We explore the music adaption of SSL with two distinctive speech-related models, data2vec1.0 and Hubert, and refer to them as music2vec and musicHuBERT, respectively. We train $12$ SSL models with 95M parameters under various pre-training configurations and systematically evaluate the MIR task performances with 13 different MIR tasks. Our findings suggest that training with music data can generally improve performance on MIR tasks, even when models are trained using paradigms designed for speech. However, we identify the limitations of such existing speech-oriented designs, especially in modelling polyphonic information. Based on the experimental results, empirical suggestions are also given for designing future musical SSL strategies and paradigms.
BTS is a music group from South Korea which has global popularity around the world. As a music group that has global popularity, BTS influence on social media is quite large, with this influence making many companies interested in making them Brand Ambassadors (BA). A company can do a sentiment analysis of the public figure they want to make BA, later this sentiment can be taken into consideration for them to be able to determine the BA of their product. Sentiment analysis can also be used to distinguish the attitude of customers, users or followers towards a brand, topic, or product with the help of their reviews. Based on this, this study will analyze the sentiments of Twitter users towards music group BTS, using the Knowledge Discovery Database (KDD) research methodology, with 5 stages namely Data Selection, Data Preprocessing, Data Transformation, Text Mining and Evaluation. By using the Support Vector Machine (SVM) algorithm with a linear kernel, this study will do 3 scenarios with the distribution of training data and testing data 90:10 in scenario 1, 80:20 in scenario 2, and 70:30 in scenario 3. Confusion Matrix is used to evaluate the performance of the algorithm used and the results show that the best performance of the model formed is in scenario 1 and scenario 2.
János Girán, Kyra Anna Girán, Dalma Ormándlaky
et al.
Consumption of energy drinks is harmful in childhood and adolescence, and its increasing popularity makes it a public health threat in this age group. Our study aimed to assess energy drink (ED) consumption and identify the context and determinants of its consumption at a Hungarian primary school. A mixed-method approach was used for the research, including a survey filled in by 157 10-15-year-old pupils and World Café workshops (WCWs) involving pupils, home-room teachers, and Parental Council representatives (N = 39). The Jamovi 2.2.5. The software was used to perform descriptive statistics and logistic regression, and a causal loop diagram was created based on the results of the WCWs. The survey results revealed that almost one-third of the pupils consumed EDs regularly, and most daily consumers drank high amounts (500ml). Most students considered ED consumption unhealthy, yet every fifth drank them. Buying breakfast on the way to school increased the odds of ED consumption almost threefold. According to the WCWs' findings, the determinants of ED consumption were embedded in two critical contextual sets; one was the need for energy and concentration boost and the Perception of high social acceptance of ED consumption. Our results suggest that interventions to reduce students' ED consumption need to include increased parents' involvement in controlling their children's screen time and encouraging them to provide breakfast at home for their children. There is also an urgent need to restrict the marketing of EDs and strictly regulate access to EDs for under-18s.
Music self-efficacy has been acknowledged as a strong predictor of successful performance among musicians and music students, but is less researched among amateur musicians. The purpose of this study is to examine the connection between music self-efficacy and learning and playing habits of amateur musicians who had preferences for online music tutorials as an informal learning technique. Two hundred and twenty-two amateur musicians completed four research questionnaires: (a) a general background questionnaire including learning and playing habits, (b) music self-efficacy, (c) self-esteem, and (d) online music tutorials users’ preferences, which distinguish between autonomy online music tutorials (fostering independent learners, incorporating context, and background) and dependent online music tutorials (offering mainly imitation options and providing little further knowledge). The results of the study supported the hypotheses, which predicted high music self-efficacy to be associated with the use of autonomy online music tutorials, co-playing, studying with a teacher, and self-esteem, and low music self-efficacy to be associated with the use of dependent online music tutorials. As music self-efficacy is crucial for students’ learning processes, we argue that music educators could refer students to optimal use of online music tutorials, thus fostering autonomous learners. Further research is required for clarifying the impact of personal characteristics and playing habits among amateur musicians on their music self-efficacy.
Carlos Hernandez-Olivan, Jorge Abadias Puyuelo, Jose R. Beltran
Deep learning models are typically evaluated to measure and compare their performance on a given task. The metrics that are commonly used to evaluate these models are standard metrics that are used for different tasks. In the field of music composition or generation, the standard metrics used in other fields have no clear meaning in terms of music theory. In this paper, we propose a subjective method to evaluate AI-based music composition systems by asking questions related to basic music principles to different levels of users based on their musical experience and knowledge. We use this method to compare state-of-the-art models for music composition with deep learning. We give the results of this evaluation method and we compare the responses of each user level for each evaluated model.