Live music provides a uniquely rich setting for studying creativity and interaction due to its spontaneous nature. The pursuit of live music agents—intelligent systems supporting real-time music performance and interaction—has captivated researchers across HCI, AI, and computer music for decades, and recent advancements in AI suggest unprecedented opportunities to evolve their design. However, the interdisciplinary nature of music has led to fragmented development across research communities, hindering effective communication and collaborative progress. In this work, we bring together perspectives from these diverse fields to map the current landscape of live music agents. Based on our analysis of 184 systems across both academic literature and video, we develop a comprehensive design space that categorizes dimensions spanning usage contexts, interactions, technologies, and ecosystems. By highlighting trends and gaps in live music agents, our design space offers researchers, designers, and musicians a structured lens to understand existing systems and shape future directions in real-time human-AI music co-creation. We release our annotated systems as a living artifact at https://live-music-agents.github.io.

Recent years have seen a surge in musical performances accompanied by generative agents. Artificial voices, timbres synthesized by neural networks, and agents that mirror or respond to human performers are rapidly taking the stage. In parallel, practitioners in human-computer interaction (HCI) and music technology have called for practice-based research that identifies the most salient affordances of these developments by examining their use in the real-world contexts of music making. To advance practice-based research on human-AI music creation, we present a longitudinal account of two months of codesign with top local jazz musicians, spanning early explorations, the identification of emerging goals, and rehearsals. Our work culminates in a public concert for a live audience of 97, featuring three pieces co-improvised with AI agents. Drawing on systems including VampNet, Somax2, and the jam_bot, each piece was tailored to the stylistic strengths of the performers and the unique strengths and limitations of each system. Through this extensive iterative process, we uncovered a wide range of design interventions, from augmenting GenAI systems with a guitar pedal to situate it in a loop-based creative practice, to enabling musicians to anticipate AI response by visually forecasting its predictions. Where musicians tended to rein in the wilder qualities of the generative systems, some audience members expected a human-AI performance to allow as much agency and spontaneity as possible. In post-concert reflection, musicians also expressed the desire to practice more which in turn could enable them to let the agency of the systems shine. They also encouraged future musicians to lean more into the uncertainty. Together, we see a unique practice emerging through this musician-AI live improv medium.

Recent multi-modal audio-language models (ALMs) excel at text-audio retrieval but struggle with frame-wise audio understanding. Prior works use temporal-aware labels or unsupervised training to improve frame-wise capabilities, but they still lack fine-grained labeling capability to pinpoint when an event occurs. While traditional sound event detection models can precisely localize events, they are limited to pre-defined categories, making them ineffective for real-world scenarios with out-of-distribution events. In this work, we introduce FLAM, an open-vocabulary contrastive audio-language model capable of localizing specific sound events. FLAM employs a memory-efficient and calibrated frame-wise objective with logit adjustment to address spurious correlations, such as event dependencies and label imbalances during training. To enable frame-wise supervision, we leverage a large-scale dataset with diverse audio events, LLM-generated captions and simulation. Experimental results and case studies demonstrate that FLAM significantly improves the open-vocabulary localization capability while maintaining strong performance in global retrieval and downstream tasks.

Jamming requires coordination, anticipation, and collaborative creativity between musicians. Current generative models of music produce expressive output but are not able to generate in an online manner, meaning simultaneously with other musicians (human or otherwise). We propose ReaLchords, an online generative model for improvising chord accompaniment to user melody. We start with an online model pretrained by maximum likelihood, and use reinforcement learning to finetune the model for online use. The finetuning objective leverages both a novel reward model that provides feedback on both harmonic and temporal coherency between melody and chord, and a divergence term that implements a novel type of distillation from a teacher model that can see the future melody. Through quantitative experiments and listening tests, we demonstrate that the resulting model adapts well to unfamiliar input and produce fitting accompaniment. ReaLchords opens the door to live jamming, as well as simultaneous co-creation in other modalities.

We propose a new method for training a supervised source separation system that aims to learn the interdependent relationships between all combinations of sources in a mixture. Rather than independently estimating each source from a mix, we reframe the source separation problem as an Orderless Neural Autoregressive Density Estimator (NADE), and estimate each source from both the mix and a random subset of the other sources. We adapt a standard source separation architecture, Demucs, with additional inputs for each individual source, in addition to the input mixture. We randomly mask these input sources during training so that the network learns the conditional dependencies between the sources. By pairing this training method with a blocked Gibbs sampling procedure at inference time, we demonstrate that the network can iteratively improve its separation performance by conditioning a source estimate on its earlier source estimates. Experiments on two source separation datasets show that training a Demucs model with an Orderless NADE approach and using Gibbs sampling (up to 512 steps) at inference time strongly outperforms a Demucs baseline that uses a standard regression loss and direct (one step) estimation of sources.

There is an increasing interest from ML and HCI communities in empowering creators with better generative models and more intuitive interfaces with which to control them. In music, ML researchers have focused on training models capable of generating pieces with increasing long-range structure and musical coherence, while HCI researchers have separately focused on designing steering interfaces that support user control and ownership. In this study, we investigate how developments in both models and user interfaces are important for empowering co-creation where the goal is to create music that communicates particular imagery or ideas (e.g., as is common for other purposeful tasks in music creation like establishing mood or creating accompanying music for another media). Our study is distinguished in that it measures communication through both composer’s self-reported experiences, and how listeners evaluate this communication through the music. In an evaluation study with 26 composers creating 100+ pieces of music and listeners providing 1000+ head-to-head comparisons, we find that more expressive models and more steerable interfaces are important and complementary ways to make a difference in composers communicating through music and supporting their creative empowerment.

Novice composers often find it difficult to go beyond common chord progressions. To make it easier for composers to experiment with radical chord choices, we built a creativity support tool, ChordRipple, which makes chord recommendations that aim to be both diverse and appropriate to the current context. Composers can use it to help select the next chord, or to replace sequences of chords in an internally consistent manner. To make such recommendations, we adapt a neural network model from natural language processing known as Word2Vec to the music domain. This model learns chord embeddings from a corpus of chord sequences, placing chords nearby when they are used in similar contexts. The learned embeddings support creative substitutions between chords, and also exhibit topological properties that correspond to musical structure. For example, the major and minor chords are both arranged in the latent space in shapes corresponding to the circle-of-fifths. Our structured observations with 14 music students show that the tool helped them explore a wider palette of chords, and to make "big jumps in just a few chords". It gave them "new ideas of ways to move forward in the piece", not just on a chord-to-chord level but also between phrases. Our controlled studies with 9 more music students show that more adventurous chords are adopted when composing with ChordRipple.

Typical synthesizers only provide controls to the low-level parameters of sound-synthesis, such as wave-shapes or filter envelopes. In contrast, composers often want to adjust and express higher-level qualities, such as how "scary" or "steady" sounds are perceived to be. We develop a system which allows users to directly control abstract, high-level qualities of sounds. To do this, our system learns functions that map from synthesizer control settings to perceived levels of high-level qualities. Given these functions, our system can generate high-level knobs that directly adjust sounds to have more or less of those qualities. We model the functions mapping from control-parameters to the degree of each high-level quality using Gaussian processes, a nonparametric Bayesian model. These models can adjust to the complexity of the function being learned, account for nonlinear interaction between control-parameters, and allow us to characterize the uncertainty about the functions being learned. By tracking uncertainty about the functions being learned, we can use active learning to quickly calibrate the tool, by querying the user about the sounds the system expects to most improve its performance. We show through simulations that this model-based active learning approach learns high-level knobs on certain classes of target concepts faster than several baselines, and give examples of the resulting automatically- constructed knobs which adjust levels of non-linear, high- level concepts.