Papers › Identity-Preserving Text-to-Video Generation by Frequency Decomposition

Identity-Preserving Text-to-Video Generation by Frequency Decomposition

26 Nov 2024CVPR 2025 1arXiv:2411.17440archive 2025-07-28

Shenghai Yuan, Jinfa Huang, Xianyi He, Yunyuan Ge, Yujun Shi, Liuhan Chen, Jiebo Luo, Li Yuan

Identity-preserving text-to-video (IPT2V) generation aims to create high-fidelity videos with consistent human identity. It is an important task in video generation but remains an open problem for generative models. This paper pushes the technical frontier of IPT2V in two directions that have not been resolved in literature: (1) A tuning-free pipeline without tedious case-by-case finetuning, and (2) A frequency-aware heuristic identity-preserving DiT-based control scheme. We propose ConsisID, a tuning-free DiT-based controllable IPT2V model to keep human identity consistent in the generated video. Inspired by prior findings in frequency analysis of diffusion transformers, it employs identity-control signals in the frequency domain, where facial features can be decomposed into low-frequency global features and high-frequency intrinsic features. First, from a low-frequency perspective, we introduce a global facial extractor, which encodes reference images and facial key points into a latent space, generating features enriched with low-frequency information. These features are then integrated into shallow layers of the network to alleviate training challenges associated with DiT. Second, from a high-frequency perspective, we design a local facial extractor to capture high-frequency details and inject them into transformer blocks, enhancing the model's ability to preserve fine-grained features. We propose a hierarchical training strategy to leverage frequency information for identity preservation, transforming a vanilla pre-trained video generation model into an IPT2V model. Extensive experiments demonstrate that our frequency-aware heuristic scheme provides an optimal control solution for DiT-based models. Thanks to this scheme, our ConsisID generates high-quality, identity-preserving videos, making strides towards more effective IPT2V. Code: https://github.com/PKU-YuanGroup/ConsisID.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

PKU-YuanGroup/ConsisID officialmentioned in papermentioned on GitHubpytorchApache-2.0 report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Human-Domain Subject-to-VideoImage to Video GenerationOpen-Domain Subject-to-VideoText-to-Video GenerationVideo Generation

Datasets

Introduced by this paper, per the archive.

ConsisID-preview-Data

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 Aesthetics 0.4460 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 FaceSim 0.4010 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 GmeScore 0.6620 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 Motion 0.4160 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 NaturalScore 0.7906 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 NexusScore 0.4592 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 Total Score 0.5446 #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Kling 1.6 Venue Close-Source #1 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 Aesthetics 0.4687 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 FaceSim 0.3080 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 GmeScore 0.6921 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 Motion 0.2470 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 NaturalScore 0.6979 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 NexusScore 0.4541 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 Total Score 0.4888 #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Pika 2.1 Venue Close-Source #6 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 Aesthetics 0.4147 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 FaceSim 0.3511 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 GmeScore 0.6757 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 Motion 0.1352 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 NaturalScore 0.7144 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 NexusScore 0.4355 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 Total Score 0.4759 #8 of 10 Archive leaderboard report
Open-Domain Subject-to-Video OpenS2V-Eval Vidu 2.0 Venue Close-Source #8 of 10 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Diffusion

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections