arXiv subjects

Meiguang Jin

Publications and source records attributed to Meiguang Jin.

16 recordsLinked to original sources

arXiv2026

TBDub: Production-Oriented Visual Dubbing

Visual dubbing must synchronize mouth motion with replacement speech while preserving identity, appearance, and temporal consistency. Although X-Dub provides a strong mask-free video-editing baseline, its application to livestream and generated-video content reveals limitations in production-domain robustness, temporal and motion stability, identity and oral-detail preservation, and inference efficiency. We present \textbf{TBDub}, a production-oriented extension of X-Dub that combines task-adaptive post-training with task-aware few-step distillation. Post-training adapts the video DiT using production-domain data, production-specific conditioning and filtering, and enhanced audio features to obtain a 30-step Teacher. Distillation adapts DMD/DMD2 to conditional video editing and compresses the Teacher into a two-step Student. On 38 TalkVid clips, the Teacher improves all eight reported reconstruction, perceptual, identity, and synchronization metrics over X-Dub. In the MOS evaluation, it improves lip-sync consistency, identity consistency, and visual quality over X-Dub by 0.14, 0.95, and 0.90 points, while the Student achieves the highest lip-sync and visual-quality scores and remains close to the Teacher in identity consistency. In paired end-to-end generation timing from the first VAE encode through the final VAE decode on a single NVIDIA H20 GPU at $512\times512$, the Student reaches 7.13 effective FPS and reduces total latency by $13.93\times$; the DiT stage alone is accelerated by $42.49\times$. The Student largely retains the Teacher's generation quality and audiovisual synchronization. The code is available on GitHub at \https://github.com/TaoLiveAIGC/TBDub, and the 30-step Teacher and two-step Student weights are available on Hugging Face at https://huggingface.co/TaoLiveAIGC/TBDub.

cs.CV↗

arXiv2026

TLive-Omni: An Omni-Modal Understanding Model for E-Commerce Live Streaming

E-commerce live streaming requires omni-modal understanding of noisy, temporally extended streams, where product facts are distributed across speech, video frames, product images, overlaid text, and user queries. We present TLive-Omni, an omni-modal understanding model tailored to live-commerce scenarios. It maps image, video, audio, and text inputs into a unified representation space. For long-form live streaming analysis, we introduce Per-vGrid, a timestamped token organization that groups each video grid with its temporally corresponding audio within explicit boundary tokens to facilitate temporal alignment. We design a three-stage supervised training recipe that progressively develops live-commerce understanding, from omni-modal perception to instruction-following responses. We then propose Faithful-RFT, a reinforcement fine-tuning stage that further improves answer faithfulness and expression quality while meeting real-time demands, scoring final responses directly with task-verifiable feedback rather than optimizing for reasoning-style exploration during rollout. Moreover, TLive-Omni is supported by a scenario-oriented atomic capability taxonomy and a compact data production engine that converts live-commerce audio, image, and video streams into training signals for speech recognition, speaker analysis, product visual grounding, text recognition, temporal grounding, video dense caption, and omni-modal QA, etc. For scalable training, a synchronized length-grouped sampler reduces padding while preserving comparable workloads across workers, while a lightweight dynamic sampling strategy regenerates rollout groups with near-zero reward variance to maintain meaningful relative advantages for GRPO. Experiments on e-commerce live streaming benchmarks demonstrate strong performance across live-commerce domain tasks, together with excellent generalization on general benchmarks.

cs.AI↗

arXiv2026

Training Agents to Evolve with Their Harness: TaoLive Digital Avatar Agent Technical Report

AI-powered digital avatar streamers must answer product questions, engage viewers, and execute marketing strategies in real time, demanding low latency, frequent strategy updates, and accurate yet effective responses. Evolvable Harnesses, whose Skills, Hooks, prompts, and tools can be updated independently of model weights, enable rapid iteration but expose a trade-off: large models adapt zero-shot yet are too slow, whereas compact models meet latency targets but overfit to fixed Harness configurations. We propose Harness-Aware Training (HAT), which trains compact models to adapt to changing Harnesses. Its key component, Harness-State Augmentation (HSA), applies task-preserving transformations to Skill identifiers and content, tool schemas, prompt structures, and Hook functions. Training proceeds in three stages: HSA-SFT learns reasoning and tool use from strong-model trajectories across diverse environments; General On-Policy Distillation restores generalization lost during SFT; and HSA-RL improves robustness to changing Harnesses through reinforcement learning in augmented environments. Across four evaluation sets, HAT achieves 94.8 on Live-Stream QA (base: 80.3; strongest general LLM: 93.0) and 94.6 on Harness-Variant QA (base: 75.4). Unlike Fixed-Harness SFT, which lowers IFEval by 7.7 points from the base model, HAT avoids this regression and reaches 83.5. On one NVIDIA H20 GPU, the optimized system delivers P50 and P95 latencies of 3.4 s and 8.1 s. Deployed in Taobao Live's digital-avatar service, it also yields positive online A/B test results for GMV and item-page views.

cs.CL↗

arXiv2026

AptAvatar: Fast and Vivid Long-Form Audio-Driven Video Generation for Production-Ready Avatars

Production-ready audio-driven avatar generation requires efficient inference without sacrificing fidelity or motion expressiveness. However, existing acceleration methods often compromise quality through restrictive architectural choices, such as causal attention and short temporal horizons, or by reducing model capacity and resolution. Without such compromises, we propose AptAvatar, a 14B-parameter long-form audio-driven avatar generation framework that delivers fast and expressive inference. For efficiency in production-level applications, AptAvatar addresses the extreme two-step generation challenge. To bridge the gap between the multi-step teacher model and the two-step student model, we introduce Endpoint-Anchored Distribution Distillation. It augments vanilla distribution matching with a dedicated Anchor Score Estimator trained on the trajectory-endpoint distribution defined from a frozen pretrained 4-step bridge generator. This provides an attainable endpoint-level anchor for the evolving two-step student. To improve long-horizon consistency, we further introduce Self-Generated History Replay, which reuses cached outputs from earlier generator checkpoints as history conditions during chunk-wise training. This approximates inference-time conditioning on self-generated histories without costly online rollouts, mitigating quality degradation from accumulated history errors. Extensive experiments demonstrate that AptAvatar generates vivid 720p long-form avatar videos with only 2 NFEs, achieving a 60x speedup while preserving visual fidelity and long-horizon identity. Code is available at https://github.com/TaoLiveAIGC/AptAvatar

cs.CV↗

arXiv2026

TaoMate: Anchor-Guided Memory Bridging Evolving and Reference States for Real-Time Audio-Video Digital Human Generation

Real-time long-form digital-human generation relies on causal models to extend audio-visual content while preserving subject appearance and audio-video synchronization across successive segments. A bounded cache retains local motion and phonetic context but discards older evidence, whereas attending to the complete generated history is computationally expensive and can propagate accumulated errors. We present \method, an anchor-guided persistent-memory framework for few-step joint audio-video generation. The framework preserves an immutable visual anchor, compresses completed video and audio blocks into fixed-capacity dynamic states, and retrieves those states through modality-specific residual attention without extending the active cache. A reference-aware modulation method additionally conditions video features on dynamic and anchor appearance statistics. Anchor-preserving causal-context distillation varies rollout horizon, prefix provenance, and cache-history reliability while keeping the immutable visual anchor unperturbed. By separating persistent memory from stage-local denoising dependencies, \method further admits stage-parallel execution across blocks, accelerating autoregressive inference without pipeline-specific retraining. We evaluate long-form video continuations with appearance, temporal, synchronization, facial, and speech diagnostics. Results show that \method preserves stable appearance across prompt-conditioned segments and strong audio-visual synchronization under autoregressive generation. Our project page is https://taoliveaigc.github.io/TaoMate.

cs.CV↗

arXiv2026

OmniScope: Modality-Decoupled Token Compression for Omnimodal Large Language Models

Existing token compression methods for omnimodal large language models typically rely on one modality to determine what to retain in the other. We show that this assumption often breaks down: for the same query, audio and video relevance often peaks at different moments. This cross-modal salience mismatch makes unidirectional guidance prone to discarding answer-critical cues under aggressive compression. We propose OmniScope, a training-free token compression framework that uses the query as a shared semantic anchor while estimating relevance separately for audio and video. OmniScope allocates modality-specific token budgets, prunes visual tokens with an anchor-delta strategy that preserves both global context and temporal changes, and merges audio tokens within each second to reduce redundancy while maintaining temporal continuity. Across four audio-video benchmarks and two Qwen2.5-Omni model scales, OmniScope achieves the best average accuracy across all compression settings. At 25% overall token retention, it delivers up to 3.53x prefill speedup and more than 15% GPU memory reduction, with only a 0.35-point drop in average accuracy. These results suggest a simple design principle for OmniLLM inference: share the query across modalities, but not the salience estimates. The code is available at https://github.com/MAC-AutoML/OmniScope.

cs.CV↗

arXiv2026

Attention Grounded Enhancement for Visual Document Retrieval

Visual document retrieval requires understanding heterogeneous and multi-modal content to satisfy implicit information needs. Recent advances use screenshot-based document encoding with fine-grained late interaction to encode holistic information and capture nuanced alignments, significantly improving retrieval performance. However, retrievers are still trained with coarse global relevance labels, without revealing which regions support the match. As a result, retrievers tend to rely on surface-level cues and struggle to capture implicit semantic connections, hindering their ability to handle non-extractive queries.To improve fine-grained relevance modeling, we propose a Attention-Grounded REtriever Enhancement (AGREE) framework. AGREE leverages cross-modal attention from multimodal large language models (MLLMs) as proxy supervision to guide the retriever in identifying relevant document regions. Specifically, AGREE extracts attention maps from the MLLM that highlight which document regions are attended to based on the query. These attention scores serve as local, region-level relevance signals. During training, AGREE combines local signals with the global document-level relevance label to jointly optimize the retriever. This dual-level supervision enables the model to learn not only whether documents match, but also which content drives relevance. Experiments on the challenging visual document retrieval benchmark, ViDoRe V2, show that AGREE significantly outperforms the global-supervision-only baseline by 12.82\% and 5.03\% in terms of average nDCG@1 and nDCG@5. Quantitative and qualitative analyses further demonstrate that AGREE promotes deeper alignment between query terms and document regions, moving beyond surface-level matching toward more accurate and interpretable retrieval. Our code is available at: https://github.com/VickiCui/AGREE.

cs.IR↗

arXiv2026

CoInteract: Physically-Consistent Human-Object Interaction Video Synthesis via Spatially-Structured Co-Generation

Synthesizing human--object interaction (HOI) videos has broad practical value in e-commerce, digital advertising, and virtual marketing. However, current diffusion models, despite their photorealistic rendering capability, still frequently fail on (i) the structural stability of sensitive regions such as hands and faces and (ii) physically plausible contact (e.g., avoiding hand--object interpenetration). We present CoInteract, an end-to-end framework for HOI video synthesis conditioned on a person reference image, a product reference image, text prompts, and speech audio. CoInteract introduces two complementary designs embedded into a Diffusion Transformer (DiT) backbone. First, we propose a Human-Aware Mixture-of-Experts (MoE) that routes tokens to lightweight, region-specialized experts via spatially supervised routing, improving fine-grained structural fidelity with minimal parameter overhead. Second, we propose Spatially-Structured Co-Generation, a dual-stream training paradigm that jointly models an RGB appearance stream and an auxiliary HOI structure stream to inject interaction geometry priors. During training, the HOI stream attends to RGB tokens and its supervision regularizes shared backbone weights; at inference, the HOI branch is removed for zero-overhead RGB generation. Experimental results demonstrate that CoInteract significantly outperforms existing methods in structural stability, logical consistency, and interaction realism.

cs.CV↗

arXiv2025

LiveThinking: Enabling Real-Time Efficient Reasoning for AI-Powered Livestreaming via Reinforcement Learning

In AI-powered e-commerce livestreaming, digital avatars require real-time responses to drive engagement, a task for which high-latency Large Reasoning Models (LRMs) are ill-suited. We introduce LiveThinking, a practical two-stage optimization framework to bridge this gap. First, we address computational cost by distilling a 670B teacher LRM into a lightweight 30B Mixture-of-Experts (MoE) model (3B active) using Rejection Sampling Fine-Tuning (RFT). This reduces deployment overhead but preserves the teacher's verbose reasoning, causing latency. To solve this, our second stage employs reinforcement learning with Group Relative Policy Optimization (GRPO) to compress the model's reasoning path, guided by a multi-objective reward function balancing correctness, helpfulness, and brevity. LiveThinking achieves a 30-fold reduction in computational cost, enabling sub-second latency. In real-world application on Taobao Live, it improved response correctness by 3.3% and helpfulness by 21.8%. Tested by hundreds of thousands of viewers, our system led to a statistically significant increase in Gross Merchandise Volume (GMV), demonstrating its effectiveness in enhancing user experience and commercial performance in live, interactive settings.

cs.LG↗

arXiv2024

Toward Tiny and High-quality Facial Makeup with Data Amplify Learning

Contemporary makeup approaches primarily hinge on unpaired learning paradigms, yet they grapple with the challenges of inaccurate supervision (e.g., face misalignment) and sophisticated facial prompts (including face parsing, and landmark detection). These challenges prohibit low-cost deployment of facial makeup models, especially on mobile devices. To solve above problems, we propose a brand-new learning paradigm, termed "Data Amplify Learning (DAL)," alongside a compact makeup model named "TinyBeauty." The core idea of DAL lies in employing a Diffusion-based Data Amplifier (DDA) to "amplify" limited images for the model training, thereby enabling accurate pixel-to-pixel supervision with merely a handful of annotations. Two pivotal innovations in DDA facilitate the above training approach: (1) A Residual Diffusion Model (RDM) is designed to generate high-fidelity detail and circumvent the detail vanishing problem in the vanilla diffusion models; (2) A Fine-Grained Makeup Module (FGMM) is proposed to achieve precise makeup control and combination while retaining face identity. Coupled with DAL, TinyBeauty necessitates merely 80K parameters to achieve a state-of-the-art performance without intricate face prompts. Meanwhile, TinyBeauty achieves a remarkable inference speed of up to 460 fps on the iPhone 13. Extensive experiments show that DAL can produce highly competitive makeup models using only 5 image pairs.

cs.CV↗

arXiv2023

Lightweight network towards real-time image denoising on mobile devices

Deep convolutional neural networks have achieved great progress in image denoising tasks. However, their complicated architectures and heavy computational cost hinder their deployments on mobile devices. Some recent efforts in designing lightweight denoising networks focus on reducing either FLOPs (floating-point operations) or the number of parameters. However, these metrics are not directly correlated with the on-device latency. In this paper, we identify the real bottlenecks that affect the CNN-based models' run-time performance on mobile devices: memory access cost and NPU-incompatible operations, and build the model based on these. To further improve the denoising performance, the mobile-friendly attention module MFA and the model reparameterization module RepConv are proposed, which enjoy both low latency and excellent denoising performance. To this end, we propose a mobile-friendly denoising network, namely MFDNet. The experiments show that MFDNet achieves state-of-the-art performance on real-world denoising benchmarks SIDD and DND under real-time latency on mobile devices. The code and pre-trained models will be released.

cs.CV↗

arXiv2022

SepLUT: Separable Image-adaptive Lookup Tables for Real-time Image Enhancement

Image-adaptive lookup tables (LUTs) have achieved great success in real-time image enhancement tasks due to their high efficiency for modeling color transforms. However, they embed the complete transform, including the color component-independent and the component-correlated parts, into only a single type of LUTs, either 1D or 3D, in a coupled manner. This scheme raises a dilemma of improving model expressiveness or efficiency due to two factors. On the one hand, the 1D LUTs provide high computational efficiency but lack the critical capability of color components interaction. On the other, the 3D LUTs present enhanced component-correlated transform capability but suffer from heavy memory footprint, high training difficulty, and limited cell utilization. Inspired by the conventional divide-and-conquer practice in the image signal processor, we present SepLUT (separable image-adaptive lookup table) to tackle the above limitations. Specifically, we separate a single color transform into a cascade of component-independent and component-correlated sub-transforms instantiated as 1D and 3D LUTs, respectively. In this way, the capabilities of two sub-transforms can facilitate each other, where the 3D LUT complements the ability to mix up color components, and the 1D LUT redistributes the input colors to increase the cell utilization of the 3D LUT and thus enable the use of a more lightweight 3D LUT. Experiments demonstrate that the proposed method presents enhanced performance on photo retouching benchmark datasets than the current state-of-the-art and achieves real-time processing on both GPUs and CPUs.

cs.CV↗

arXiv2022

NTIRE 2022 Challenge on Efficient Super-Resolution: Methods and Results

This paper reviews the NTIRE 2022 challenge on efficient single image super-resolution with focus on the proposed solutions and results. The task of the challenge was to super-resolve an input image with a magnification factor of $\times$4 based on pairs of low and corresponding high resolution images. The aim was to design a network for single image super-resolution that achieved improvement of efficiency measured according to several metrics including runtime, parameters, FLOPs, activations, and memory consumption while at least maintaining the PSNR of 29.00dB on DIV2K validation set. IMDN is set as the baseline for efficiency measurement. The challenge had 3 tracks including the main track (runtime), sub-track one (model complexity), and sub-track two (overall performance). In the main track, the practical runtime performance of the submissions was evaluated. The rank of the teams were determined directly by the absolute value of the average runtime on the validation set and test set. In sub-track one, the number of parameters and FLOPs were considered. And the individual rankings of the two metrics were summed up to determine a final ranking in this track. In sub-track two, all of the five metrics mentioned in the description of the challenge including runtime, parameter count, FLOPs, activations, and memory consumption were considered. Similar to sub-track one, the rankings of five metrics were summed up to determine a final ranking. The challenge had 303 registered participants, and 43 teams made valid submissions. They gauge the state-of-the-art in efficient single image super-resolution.

cs.CV↗

arXiv2022

AdaInt: Learning Adaptive Intervals for 3D Lookup Tables on Real-time Image Enhancement

The 3D Lookup Table (3D LUT) is a highly-efficient tool for real-time image enhancement tasks, which models a non-linear 3D color transform by sparsely sampling it into a discretized 3D lattice. Previous works have made efforts to learn image-adaptive output color values of LUTs for flexible enhancement but neglect the importance of sampling strategy. They adopt a sub-optimal uniform sampling point allocation, limiting the expressiveness of the learned LUTs since the (tri-)linear interpolation between uniform sampling points in the LUT transform might fail to model local non-linearities of the color transform. Focusing on this problem, we present AdaInt (Adaptive Intervals Learning), a novel mechanism to achieve a more flexible sampling point allocation by adaptively learning the non-uniform sampling intervals in the 3D color space. In this way, a 3D LUT can increase its capability by conducting dense sampling in color ranges requiring highly non-linear transforms and sparse sampling for near-linear transforms. The proposed AdaInt could be implemented as a compact and efficient plug-and-play module for a 3D LUT-based method. To enable the end-to-end learning of AdaInt, we design a novel differentiable operator called AiLUT-Transform (Adaptive Interval LUT Transform) to locate input colors in the non-uniform 3D LUT and provide gradients to the sampling intervals. Experiments demonstrate that methods equipped with AdaInt can achieve state-of-the-art performance on two public benchmark datasets with a negligible overhead increase. Our source code is available at https://github.com/ImCharlesY/AdaInt.

cs.CV↗

arXiv2018

Learning to Extract a Video Sequence from a Single Motion-Blurred Image

We present a method to extract a video sequence from a single motion-blurred image. Motion-blurred images are the result of an averaging process, where instant frames are accumulated over time during the exposure of the sensor. Unfortunately, reversing this process is nontrivial. Firstly, averaging destroys the temporal ordering of the frames. Secondly, the recovery of a single frame is a blind deconvolution task, which is highly ill-posed. We present a deep learning scheme that gradually reconstructs a temporal ordering by sequentially extracting pairs of frames. Our main contribution is to introduce loss functions invariant to the temporal order. This lets a neural network choose during training what frame to output among the possible combinations. We also address the ill-posedness of deblurring by designing a network with a large receptive field and implemented via resampling to achieve a higher computational efficiency. Our proposed method can successfully retrieve sharp image sequences from a single motion blurred image and can generalize well on synthetic and real datasets captured with different cameras.

cs.CV↗

arXiv2017

Deep Mean-Shift Priors for Image Restoration

In this paper we introduce a natural image prior that directly represents a Gaussian-smoothed version of the natural image distribution. We include our prior in a formulation of image restoration as a Bayes estimator that also allows us to solve noise-blind image restoration problems. We show that the gradient of our prior corresponds to the mean-shift vector on the natural image distribution. In addition, we learn the mean-shift vector field using denoising autoencoders, and use it in a gradient descent approach to perform Bayes risk minimization. We demonstrate competitive results for noise-blind deblurring, super-resolution, and demosaicing.

cs.CV↗