Work

Toward Generation-Centric Coding: Compressing Latents representation for TI2V Synthesis

Year: 2026

Type: conference-paper

Abstract: Text-Image-to-Video (TI2V) generation can be sensitive to compression distortions in the reference image. Small pixel-space artifacts may be amplified through iterative generation, leading to temporal inconsistency and semantic drift. One contributing factor is that conventional codecs are optimized for human-oriented pixel fidelity, without explicitly accounting for the distribution shift induced in the generator’s conditioning latents. Motivated by this mismatch, we propose a generation-centric latent-domain coding framework that compresses generator-facing deep latents instead of raw pixels, and jointly optimizes rate and latent-space distortion to preserve conditioning statistics. Experiments across four bitrate levels show that our approach mitigates distortion accumulation over iterations. It improves temporal coherence and content fidelity for both short and extended sequences, consistently reducing the pixel-level discrepancy relative to videos generated with lossless conditioning.

Source: International Conference on Multimedia Retrieval

DOI: 10.1145/3805622.3810600

Authors: Jianran Liu, Wen Ji, Xiaokai Meng, Wancai Zhang, Ying Wang

Institutions: Chinese Academy of Sciences, Institute of Computing Technology, State Grid Corporation of China (China), State Grid Shanxi Electric Power Company Electric Power Research Institute, NARI Group (China)

Language: English

Topic: Generative Adversarial Networks and Image Synthesis

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

OpenAlex ID: W7164808897

FWCI:

Cites: 29

Cited by:

Related to: 10