Wenxi Chen, Ruiqi Yan, Yushen Chen, Zhikang Niu, Ziyang Ma, Xiquan Li, Yuzhe Liang, Wenhanlin, Shunshun Yin, Ming Tao, Xinsheng Wang, Xie Chen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

In this work, we introduce the first autoregressive framework for real-time, audio-driven portrait animation, a.k.a, talking head. Beyond the challenge of lengthy animation times, a critical challenge in realistic talking head generation lies in preserving the natural movement of diverse body parts. To this end, we propose Teller, the first streaming audio-driven protrait animation framework with autoregressive motion generation. Specifically, Teller first decomposes facial and body detail animation into two components: Facial Motion Latent Generation (FMLG) based on an autoregressive transfromer, and movement authenticity refinement using a Efficient Temporal Module (ETM). Concretely, FMLG employs a Residual VQ model to map the facial motion latent from the implicit keypoint-based model into discrete motion tokens, which are then temporally sliced with audio embeddings. This enables the AR tranformer to learn real-time, stream-based mappings from audio to motion. Furthermore, Teller incorporate ETM to capture finer motion details. This module ensures the physical consistency of body parts and accessories, such as neck muscles and earrings, improving the realism of these movements. Teller is designed to be efficient, surpassing the inference speed of diffusion-based models (Hallo 20.93s vs. ${\mathbf{Teller0}}{\mathbf{.92s}}$ for one second video generation), and achieves a real-time streaming performance of up to ${\mathbf{25 FPS}}$. Extensive experiments demonstrate that our method outperforms recent audio-driven portrait animation models, especially in small movements, as validated by human evaluations with a significant margin in quality and realism.

Sentiment analysis tends to use automated approaches to mine the sentiment information expressed in text, such as reviews, blogs and forum discussions. As most traditional approaches for sentiment analysis are based on supervised learning models and need many labeled corpora as their training data which are not always easily obtained, various unsupervised models based on Latent Dirichlet Allocation (LDA) have been proposed for sentiment classification. In this paper, we propose a novel probabilistic modeling framework based on LDA, called Dependency-Topic-Affects-Sentiment-LDA (DTAS) model, which drops the "bag of words" assumption and assumes that the topics of sentences in a document form a Markov chain, and the sentiment of one sentence is affected by its corresponding topic and its previous sentence's topic. We applied DTAS to reviews of books and hotels. The experiment results of sentiment classification shows that DTAS outperforms other unsupervised generative models and gets high and stable accuracy.