The art of communication beyond speech there are gestures. automatic co-speech gesture generation draws much attention in computer animation. It is a challenging task due to the diversity gestures and difficulty matching rhythm semantics corresponding speech. To address these problems, we present DiffuseStyleGesture, diffusion model based speech-driven approach. generates high-quality, speech-matched, stylized, diverse on given speeches arbitrary length. Specifically, introduce cross-local...

Speech-driven gesture generation is highly challenging due to the random jitters of human motion. In addition, there an inherent asynchronous relationship between speech and gestures. To tackle these challenges, we introduce a novel quantization-based phase-guided motion matching framework. Specifically, first present VQ-VAE module learn codebook summarize meaningful units. With each code representing unique gesture, jittering problems are alleviated effectively. We then use Levenshtein...

This paper describes the ReprGesture entry to Generation and Evaluation of Non-verbal Behaviour for Embodied Agents (GENEA) challenge 2022. The GENEA provides processed datasets performs crowdsourced evaluations compare performance different gesture generation systems. In this paper, we explore an automatic system based on multimodal representation learning. We use WavLM features audio, FastText text position rotation matrix gesture. Each modality is projected two distinct subspaces:...

The Trouble with Theater: Cinema and the Geopolitics of Medium Specificity Weihong Bao (bio) As media proliferate, mutate, commingle at accelerated unprecedented speed scale today, identity cinema haunts us, perhaps more than other moments in history, an intensified urgency on which hinges stakes field, its legitimacy, ongoing legacy. This has stimulated a resurgent interest question medium, whether terms persistent or changing specificity interaction media, their mutual transmutation...

This essay explores a critical dialogue between methods and conceptions of cultural techniques—the second wave media archaeology—and case in contemporary Chinese documentary. I examine filmmaker Mao Chenyu, who is also an organic farmer, thinker writer, film exhibitor. provides intriguing how ethnography, ecology, cosmology intertwine; art can take the form activism by redefining its boundaries exhibition space; be rethought replacing usual focus on as object with space, community, social...

Research Article| December 01 2005 From Pearl White to Rose Woo: Tracing the Vernacular Body of Nüxia in Chinese Silent Cinema, 1927-1931 Weihong Bao Search for other works by this author on: This Site Google Camera Obscura (2005) 20 (3 (60)): 193–231. https://doi.org/10.1215/02705346-20-3_60-193 Cite Icon Share Twitter Permissions Citation Bao; 1927-1931. 1 2005; doi: Download citation file: Zotero Reference Manager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex toolbar search...

This essay examines the neglected wartime Chongqing cinema by situating it in its local and simultaneously global context of exhibition. Instead reinforcing image as sheer state propaganda, I illustrate film-makers' film critics' heightened awareness multiple contexts propose to consider this a search for 'cinematic Esperanto', an aspiration toward world international language that contested universal Hollywood continuity system so bridge aesthetics audience responses register atrocity war...

Audio-driven talking face with portrait customization enhances the flexibility of avatar applications for different scenarios, such as on-line meetings, mixed reality, and data generation. Among existing methods, audio-driven swapping are typically viewed separate tasks that cascaded to achieve objective. Using state-of-the-art methods Wav2Lip SimSwap this purpose, we meet some issues: affected mouth synchronization, lost texture information, slow inference speed. To resolve these issues,...

Current talking face generation methods mainly focus on speech-lip synchronization. However, insufficient investigation the facial style leads to a lifeless and monotonous avatar. Most previous works fail imitate expressive styles from arbitrary video prompts ensure authenticity of generated video. This paper proposes an unsupervised variational transfer model (VAST) vivify neutral photo-realistic avatars. Our consists three key components: encoder that extracts representations given...

AbstractTsai Ming-liang's film aesthetics has been largely compared with European auteur film-making in style and philosophical concerns. This essay attempts to resituate Tsai's works relation his prior theatre practice deployment of popular genres that carry a dialogue theories genre intermediality. I focus on The Wayward Cloud as case study examine the three distinct modes production within film, namely, avant-garde acting, musical, hardcore pornography, showing how Tsai Ming-liang crosses...

This article revisits Miriam Hansen's theory of vernacular modernism by putting it in dialogue with wartime Chongqing propaganda film theory. Comparing key points parallel between dual interest the linguistic and sensorial function cinema theory, Bao situates latter light changing conditions for production exhibition a media ecology. She pursues how fantasy as vibratory medium nonneutral, pervasive social milieu inherits historical conceptions ether developed Chinese political philosophy...

In this article, I explore the promise and pitfalls of medium as environment by tracking twin developments environmental thinking set design in China, considering it a problematic epistemology, technology, aesthetics. treat huanjing (environment) neologism, new episteme, dispositif, mode power, taking companion that reconnects art aesthetics politics. Reconceptualizing set, design, at intersection industrial progressive education, focus on modernist propagandistic practice China 1930s ’40s,...

Current talking face generation methods mainly focus on speech-lip synchronization. However, insufficient investigation the facial style leads to a lifeless and monotonous avatar. Most previous works fail imitate expressive styles from arbitrary video prompts ensure authenticity of generated video. This paper proposes an unsupervised variational transfer model (VAST) vivify neutral photo-realistic avatars. Our consists three key components: encoder that extracts representations given...

Deep Neural Networks suffer significant performance degeneration when noisy labels corrupt latent data representations. Previous work has attempted to alleviate this problem by exploiting contrastive learning, the pair building of which is critical. However, existing methods either conduct sample-level processes and then use resultant subset construct pairs or directly perform pair-level selecting using a fixed threshold, both leading sub-optimal pairing subsequent representation learning....

These short reflections, from UC Berkeley faculty in a variety of disciplines, respond to the following question: “What does phrase ‘time-based art’ mean you? What are central stakes, conventions, challenges, and opportunities durational art contexts which you work?” Collectively, they probe wide range practices contexts, including, for example, Mexican festivals midwestern American carnivals, Syrian documentary films “image-event,” bystander recordings US police state harassment black men,...