Publications
See Google Scholar for the most up-to-date information.
2026
- Precise Video-to-Audio Generation with Cross-Modal Alignment in Latent SpaceThanh V. T. Tran, Ngoc-Son Nguyen, Luong Tran, and 4 more authorsEuropean Conference on Computer Vision, 2026
Video-to-audio (V2A) generation aims to synthesize realistic audio that is both semantically consistent with and temporally synchronized to a silent video. Despite recent progress, many methods still rely on multi-stage training, resulting in high computational costs and long runtimes, or transform visual input into text to leverage pretrained text-to-audio models, sacrificing fine-grained temporal cues. To overcome these limitations, we propose Flowley, an end-to-end, single-stage training architecture that produces soundtracks by combining visual features with textual prompts. Crucially, we introduce Progressive Soft-masked Cross-Attention, which embeds audio-visual synchronization directly within its attention mechanism, adding zero additional computational cost compared to standard attention layers. We further observe that existing V2A benchmarks lack sound-oriented descriptive captions, which can potentially degrade the quality of the synthesized audio. To remedy this, we propose SoundCap, a plug-and-play pipeline for creating detailed, sound-aware captions that guide the model. Remarkably, without integrating any pretrained audio-visual alignment modules, Flowley achieves state-of-the-art performance on VGGSound across multiple metrics. Moreover, by incorporating SoundCap, we further exceed the performance of the strongest existing close-sourced methods in terms of audio quality in the zero-shot setting.
- DiFlow-TTS: Compact and Low-Latency Zero-Shot Text-to-Speech with Factorized Discrete Flow MatchingNgoc-Son Nguyen, Thanh V. T. Tran, Hieu-Nghia Huynh-Nguyen, and 2 more authorsInterspeech, 2026
Zero-shot text-to-speech (TTS) has made significant progress in replicating unseen voices, yet balancing generation quality and inference efficiency remains challenging. Autoregressive models suffer from high latency, while diffusion-based approaches are constrained by training-time configurations. Moreover, most flow-based methods operate in continuous space, which introduces optimization challenges because continuous token spaces are inherently more complex than discrete ones. To address this limitation, we propose DiFlow-TTS, a zero-shot TTS framework based on discrete flow matching. The model consists of a deterministic Phoneme-Content Mapper for linguistic modeling and a Factorized Discrete Flow Denoiser that jointly generates prosody and acoustic token streams. Experimental results demonstrate the effectiveness of our approach across multiple evaluation metrics.
- DiFlowDubber: Discrete Flow Matching for Automated Video Dubbing via Cross-Modal Alignment and SynchronizationComputer Vision and Pattern Recognition Findings, 2026
Video dubbing requires content accuracy, expressive prosody, high-quality acoustics, and precise lip synchronization, yet existing approaches struggle on all four fronts. To address these issues, we propose DiFlowDubber, the first video dubbing framework built upon a discrete flow matching backbone with a novel two-stage training strategy. In the first stage, a zero-shot text-to-speech (TTS) system is pre-trained on large-scale corpora, where a deterministic architecture captures linguistic structures, and the Discrete Flow-based Prosody-Acoustic (DFPA) module models expressive prosody and realistic acoustic characteristics. In the second stage, we propose the Content-Consistent Temporal Adaptation (CCTA) to transfer TTS knowledge to the dubbing domain: its Synchronizer enforces cross-modal alignment for lip-synchronized speech. Complementarily, the Face-to-Prosody Mapper (FaPro) conditions prosody on facial expressions, whose outputs are then fused with those of the Synchronizer to construct rich, fine-grained multimodal embeddings that capture prosody-content correlations, guiding the DFPA to generate expressive prosody and acoustic tokens for content-consistent speech. Experiments on two benchmark datasets demonstrate that DiFlowDubber outperforms prior methods across multiple evaluation metrics.
2025
- Speech Reconstruction from Silent Videos via Acoustic-Semantic Decomposed ModelingLong-Khanh Pham, Thanh V. T. Tran, Minh-Tan Pham, and 1 more authorInterspeech, 2025
Lip-to-speech (L2S) synthesis, which reconstructs speech from visual cues, faces challenges in accuracy and naturalness due to limited supervision in capturing linguistic content, accents, and prosody. In this paper, we propose RESOUND, a novel L2S system that generates intelligible and expressive speech from silent talking face videos. Leveraging source-filter theory, our method involves two components: an acoustic path to predict prosody and a semantic path to extract linguistic features. This separation simplifies learning, allowing independent optimization of each representation. Additionally, we enhance performance by integrating speech units, a proven unsupervised speech representation technique, into waveform generation alongside mel-spectrograms. This allows RESOUND to synthesize prosodic speech while preserving content and speaker identity. Experiments conducted on two standard L2S benchmarks confirm the effectiveness of the proposed method across various metrics.
- Effective Context Modeling Framework for Emotion Recognition in ConversationsCuong Tran Van*, Thanh V. T. Tran*, Van Nguyen, and 1 more authorInternational Conference on Acoustics, Speech, and Signal Processing, 2025
Emotion Recognition in Conversations (ERC) facilitates a deeper understanding of the emotions conveyed by speakers in each utterance within a conversation. Recently, Graph Neural Networks (GNNs) have demonstrated their strengths in capturing data relationships, particularly in contextual information modeling and multimodal fusion. However, existing methods often struggle to fully capture the complex interactions between multiple modalities and conversational context, limiting their expressiveness. To overcome these limitations, we propose ConxGNN, a novel GNN-based framework designed to capture contextual information in conversations. ConxGNN features two key parallel modules: a multi-scale heterogeneous graph that captures the diverse effects of utterances on emotional changes, and a hypergraph that models the multivariate relationships among modalities and utterances. The outputs from these modules are integrated into a fusion layer, where a cross-modal attention mechanism is applied to produce a contextually enriched representation. Additionally, ConxGNN tackles the challenge of recognizing minority or semantically similar emotion classes by incorporating a re-weighting scheme into the loss functions. Experimental results on the IEMOCAP and MELD benchmark datasets demonstrate the effectiveness of our method, achieving state-of-the-art performance compared to previous baselines.
- GROOT: Effective Design of Biological Sequences with Limited Experimental DataThanh V. T. Tran*, Nhat Khang Ngo*, Viet Anh Nguyen, and 1 more authorConference on Knowledge Discovery and Data Mining, 2025
Latent space optimization (LSO) is a powerful method for designing discrete, high-dimensional biological sequences that maximize expensive black-box functions, such as wet lab experiments. This is accomplished by learning a latent space from available data and using a surrogate model fΦ to guide optimization algorithms toward optimal outputs. However, existing methods struggle when labeled data is limited, as training fΦ with few labeled data points can lead to subpar outputs, offering no advantage over the training data itself. We address this challenge by introducing GROOT, a GRaph-based Latent SmOOThing for Biological Sequence Optimization. In particular, GROOT generates pseudo-labels for neighbors sampled around the training latent embeddings. These pseudo-labels are then refined and smoothed by Label Propagation. Additionally, we theoretically and empirically justify our approach, demonstrate GROOT’s ability to extrapolate to regions beyond the training set while maintaining reliability within an upper bound of their expected distances from the training regions. We evaluate GROOT on various biological sequence design tasks, including protein optimization (GFP and AAV) and three tasks with exact oracles from Design-Bench. The results demonstrate that GROOT equalizes and surpasses existing methods without requiring access to black-box oracles or vast amounts of labeled data, highlighting its practicality and effectiveness.
- LatentDE: Latent-based Directed Evolution for Protein Sequence DesignThanh V. T. Tran, Nhat Khang Ngo, Viet Thanh Duy Nguyen, and 1 more authorMachine Learning: Science and Technology, 2025
Directed evolution (DE) has been the most effective method for protein engineering that optimizes biological functionalities through a resource-intensive process of screening or selecting among a vast range of mutations. To mitigate this extensive procedure, recent advancements in machine learning-guided methodologies center around the establishment of a surrogate sequence-function model. In this paper, we propose latent-based DE (LDE), an evolutionary algorithm designed to prioritize the exploration of high-fitness mutants in the latent space. At its core, LDE is a regularized variational autoencoder (VAE), harnessing the capabilities of the state-of-the-art protein language model, ESM-2, to construct a meaningful latent space of sequences. From this encoded representation, we present a novel approach for efficient traversal on the fitness landscape, employing a combination of gradient-based methods and DE. Experimental evaluations conducted on eight protein sequence design tasks demonstrate the superior performance of our proposed LDE over previous baseline algorithms.
2024
- Protein Design by Directed Evolution Guided by Large Language ModelsThanh V. T. Tran and Truong-Son HyIEEE Transactions on Evolutionary Computation, 2024
Directed evolution, a strategy for protein engineering, optimizes protein properties (i.e., fitness) by a rigorous and resource-intensive process of screening or selecting among a vast range of mutations. By conducting an in-silico screening of sequence properties, machine learning-guided directed evolution (MLDE) can expedite the optimization process and alleviate the experimental workload. In this work, we propose a general MLDE framework in which we apply recent advancements of deep learning in protein representation learning and protein property prediction to accelerate the searching and optimization processes. In particular, we introduce an optimization pipeline that utilizes the large language models (LLMs) to pinpoint the mutation hotspots in the sequence and then suggest replacements to improve the overall fitness. Our experiments have shown the superior efficiency and efficacy of our proposed framework in the conditional protein generation, in comparison with the other state-of-the-art baseline algorithms. We expect this work will shed a new light on not only protein engineering but also on solving the combinatorial problems using the data-driven methods.