Image credit: Poespin Concept Art CGI rendering by Cory and Reraner
As an Art Papers feature at SIGGRAPH 2025, the creators of “PoeSpin: A Human–AI Dance to Poetry System for Movement-Based Verse Generation” are preparing to showcase a project that blends movement, language, and machine learning into a singular expressive form. Ahead of their presentation in Vancouver, we caught up with Yihua Li, Wendy Li, and Yetong Xin to learn more about how “PoeSpin” transforms pole dance into poetic verse, challenges cultural assumptions, and reimagines the possibilities of human–AI collaboration.
SIGGRAPH: Share some background about your SIGGRAPH 2025 Art Paper, “PoeSpin: A Human–AI Dance to Poetry System for Movement-Based Verse Generation.” What inspired this project?
Yihua Li (Cory): Hi, I am Cory, the leading AI artist and researcher of “PoeSpin.” Wendy is the pole dancer we invited, Archy is our co-author who gave us the knowledge of English literature, and Yetong is the CGI artist who supported this project’s dancing avatar visualization.
This project came from a blend of my fascination with computational linguistics and embodied artistic expressions. Observing pole dance closely, I recognized its powerful yet undervalued potential for narrative and poetic exploration. By using AI as a mediator, “PoeSpin” captures pole dance movements to “write” poetry, challenging societal biases and providing a fresh perspective on dance as narrative art. Our approach integrates computational techniques, including dimensionality reduction and AI-driven semantic mapping, to create emotionally resonant poetry from physical motion. I am glad that we can share our findings at SIGGRAPH 2025.
Wendy Li (WL): As a dancer, UI/UX designer, and someone deeply invested in the intersection of art, design, and technology, “PoeSpin” emerged from my desire to explore and deepen the expressive potential of pole dance. For me, pole dance is more than just physical movement — it serves as a profound medium for self-expression, allowing the body to communicate emotion and narrative. Through “PoeSpin”, I sought to bring this art form into a new context by integrating AI, thereby illustrating how pole dance can be understood not only as a physical practice but as a legitimate and meaningful artistic expression. Additionally, this project provides an opportunity to advocate for the pole dance community, challenging longstanding societal biases and offering a platform for its recognition. My aim is to demonstrate that pole dance, much like any other art form, holds the capacity for deep emotional and narrative exploration, deserving of respect and recognition both within the artistic community and in broader societal contexts.
SIGGRAPH: What were some of the biggest technical or conceptual challenges you encountered while developing “PoeSpin”, especially in preserving the emotional depth of both dance and poetry?
Cory: Beyond translating movement into poetry, one major technical challenge was visualizing the underlying data in an expressive and scalable way. I explored t-SNE extensively and used Blender to create 3D text particle visualizations — but due to software and workflow limits, I could only render small batches at a time. Our proposed FFT-based semantic mapping also lacks a more intuitive visualization method at this stage. I’m eager to collaborate with computer visualization scientists to push this further and make these abstract spaces more perceptible and intuitive for broader audiences.
SIGGRAPH: Pole dance often faces cultural bias — how did you approach using this medium not just as an input, but as a statement on artistic legitimacy and inclusivity?
WL: Through “PoeSpin”, I have initiated a conversation that transcends the boundaries of self-expression and external perception. This project serves as a dialogue between my embodied movements and the machine’s response, exploring how organic, human motion interacts with AI-driven processes. The collision of emotion and technology, and the negotiation between self-expression and the external gaze, create a dynamic space for inquiry. How can we express the nuanced connection between these two dialogues — the intimate, personal expression and the broader, societal recognition? “PoeSpin” offers a potential framework for this interaction, where the machine becomes not just a tool, but a partner in the exploration of movement, emotion, and artistic legitimacy. This work challenges the conventional perception of pole dance, proposing it as an art form worthy of re-examination and deeper recognition.
SIGGRAPH: In your view, what does “PoeSpin” reveal about the evolving relationship between human embodiment and machine creativity?
Cory: “PoeSpin” touches on a deeper question central to embodied intelligence — not just how machines can learn to sense our world, but how we, as humans, might begin to perceive the machine’s own abstract, mathematical world. Through movement, we explored AI not just as a tool, but as a parallel intelligence with its own sensory logic.
This led me to re-think the role of avatars — not merely as cool digital visuals or stand-ins, but as mediators between human embodiment and computational perception. “PoeSpin” offered a glimpse of how virtual bodies can traverse semantic terrain, not just mimic human motion. This realization shaped my ongoing work with VTubers and embodied agents — not just to replicate human behavior, but to explore expressions beyond human, tapping into the alien poetics of machine cognition.
Yetong Xin (YX): From my perspective, “PoeSpin” reveals the deep connection between human embodiment and machine creativity, especially from the visual design standpoint. As the visual designer responsible for this project, I believe “PoeSpin” romantically links human embodiment with machine creativity by transforming the ancient form of dance into a medium through which emotion is expressed.
Dance, as one of the oldest forms of human communication, has always conveyed emotions through bodily movements. It is not just physical movement, but the embodiment of emotions, where every gesture and posture communicates inner feelings and stories. “PoeSpin” cleverly takes this human method of emotional expression and transforms it into a language that machines can understand and create. By integrating AI, the machine translates what would otherwise be abstract data and logic into a sensory experience, almost as if it too can “understand” dance and feel the emotions of the body.
This innovative approach breaks down the traditional boundaries between humans and machines. The machine is no longer just a tool; through understanding and capturing the dynamics of dance, it transforms these movements into emotional language, demonstrating perception abilities similar to human emotional communication. “PoeSpin”, therefore, is not just a technical presentation — it’s a new form of sensory resonance that connects machine logic and human emotions through the embodied expression of dance, blending art and technology in an entirely new way.
Through “PoeSpin”, I believe we can not only see how machines “learn” human expressions but also feel the emotional resonance between humans and machines.
WL: From a pole dancer’s perspective, “PoeSpin” reveals how AI can enhance the emotional depth of embodied movement. Pole dance, with its rich expressive potential, becomes an even more powerful medium when interacting with AI, which doesn’t just replicate movement but adds layers of meaning and emotion.
This collaboration allows the performer to connect with the audience in new ways, where AI deepens the emotional resonance of the dance, pushing the boundaries of traditional performance and transforming pole dance into a more interactive and expressive art form.
SIGGRAPH: Looking ahead, how do you envision systems like “PoeSpin” expanding or influencing future cross-modal collaborations between AI and performance art?
Cory: Reflecting on “PoeSpin”, I realized we unintentionally tried a novel pathway for language generation — one that fuses embodied interaction with non-autoregressive semantic sampling. t-SNE, a nonlinear dimensionality reduction technique, served as our “semantic map,” and the dancer’s body traversed this map like an agent exploring a latent space. Each gesture became a soft projection of a potential token — resembling how diffusion models reconstruct images in latent space, except our generation was movement-triggered, not autoregressive. In this way, “PoeSpin” becomes a multimodal language generation system where the body dances through language’s latent terrain.
The stage of performance art is no longer just physical — it expands into semantic space, into the machine’s representational imagination. I’m excited by the future work that blends embodied movement with semantic systems, where language becomes not only something to read, but something to inhabit, move through, and feel.
YX: From a visual art perspective, I see systems like “PoeSpin” as a transformative force in the future of AI and performance art collaborations. These systems allow for real-time interaction between human movement and dynamic, immersive visuals, where nonlinear storytelling and interactive performances thrive. This fusion of human expression with AI creativity will challenge traditional art forms and open new possibilities for emotionally resonant and ever-evolving art. Ultimately, AI and performance art will create a new, collaborative frontier that expands how we experience and create art.
WL: “PoeSpin” demonstrates the transformative potential of integrating AI with performance art, particularly from a dancer’s perspective. It bridges the gap between human movement and machine interpretation, where the dancer’s body becomes an active agent interacting with AI-driven processes. By utilizing techniques such as t-SNE, which maps semantic space, each movement triggers real-time, dynamic language generation, creating an immersive and evolving experience. This fusion of embodied expression with AI technology challenges traditional boundaries of performance and opens new possibilities for interactive, emotionally resonant art. As a dancer, systems like “PoeSpin” not only expand how we create art but also redefine how we experience movement and language, enabling deeper connections between the performer, the technology, and the audience.
SIGGRAPH: What do you hope SIGGRAPH 2025 participants take away from your Art Papers presentation?
Cory: I’m eager to speak with curators and gallerists interested in AI-native or movement-based art. Through this work, I’ve come to understand that pole dance has repeatedly sought Olympic recognition — alongside disciplines like gymnastics — yet continues to be excluded due to deep-seated cultural biases. With “PoeSpin”, I hope to invite a deeper reconsideration of the expressive power of pole dance — not just as something to be looked at, but something to be read.
YX: As a visual artist, I hope SIGGRAPH 2025 participants take away an understanding of how AI can enhance and expand the expressive potential of performance art, particularly through the lens of embodied movement. Through “PoeSpin,” I want to highlight how dance, when combined with AI, can reframe traditional art forms, offering new ways to engage with narrative and emotion. I also hope to spark conversations about the cultural legitimacy of performance art like pole dance, which is often overlooked despite its deep artistic value. Ultimately, I hope to inspire more cross-modal collaborations that bring together technology, movement, and visual art in meaningful ways.
WL: Beyond exploring AI-native and movement-based art, I hope “PoeSpin” contributes to broader recognition of pole dance as a legitimate artistic medium. Pole dance has long been marginalized due to cultural biases, despite its profound expressive capabilities. Through “PoeSpin”, I aim to cultivate deeper understanding of pole dance — not simply as performance to be observed, but as meaningful art that invites interpretation and appreciation across multiple dimensions. I’m also interested in examining how AI and visual art can offer innovative contributions to underrepresented communities, both within performance art and society more broadly. As a dancer, my ultimate hope is that audiences will develop genuine appreciation for pole dance as an emotionally resonant and legitimate form of artistic expression that extends far beyond its technological and visual elements.
Don’t miss your chance to experience “PoeSpin” and other projects redefining the boundaries of art, performance, and technology. Register for SIGGRAPH 2025 today and be part of the global conversation shaping the future of human–AI creativity.
Yihua Li (Cory) is a media artist, creative technologist, and founder of Echuu, a 3D AI VTubing platform. As an NYU ITP graduate, his practice centers on recreating life through computational means—developing AI agents, artificial life systems, and embodied digital beings that explore new forms of consciousness and creativity.
His work has been showcased at SIGGRAPH (Long Art Paper 2025), NeurIPS Creative AI Track, CVPR AI Art Gallery, and ACM Creativity & Cognition. He has received honors including the MUSE Design Gold Award, LA A+D Museum Generative AI Award, and recognition from NYC ARTECHOUSE and Deji Art Museum. He serves as a reviewer for the 2025 NeurIPS Creative AI Track.
Yetong Xin is a director and 3D artist specializing in 3D visuals, motion design, AIGC, XR technologies, and human-computer interaction. Her work integrates computer graphics (CG), AI-generated content, and immersive technology, spanning animation, interactive installations, architectural projections, and new media experiences.
She has received recognition at NeurIPS and CHI, along with prestigious awards such as the IDA Design Award, MUSE Design Award, and A+D Museum Design Award. Her work has been exhibited across Asia, Europe, and North America, featured in over six countries worldwide.
Wendy Li is a pole dancer and designer who bridges movement and technology through UI/UX, product design, and interaction design. She approaches design as a means of shaping human experiences and challenging societal perceptions, exploring how dance, technology, and artistic expression can intersect to amplify marginalized voices. Wendy’s work demonstrates how performance and design can create deeper human connections, crafting emotionally resonant experiences that engage with digital spaces while addressing broader societal issues through art and technology.
