EchoStyle: Unlocking High-Fidelity Video Stylization with Reverse Data Synthesis

Huaqiu Li *, Jiahao Wang *, Sijia Cai †, Hualian Sheng, Bing Deng, Jieping Ye, Wenhan Luo † · 2026

Read EchoStyle: Unlocking High-Fidelity Video Stylization with Reverse Data Synthesis by Huaqiu Li *, Jiahao Wang *, Sijia Cai †, Hualian Sheng, Bing Deng, Jieping Ye, Wenhan Luo † (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-25465.txt.

Introduction

Restyling a video in a painterly look breaks easily: styles leak into the wrong objects, motion warps, and long clips drift. Huaqiu Li and colleagues present EchoStyle in 2026, a text driven framework that restyles videos of any length. Rather than condition on a reference image, a video to video design fuses the original content with the requested style described in words. The training data problem is attacked backwards, by synthesising stylised pairs automatically.

That reverse pipeline produced a dataset of twenty thousand high quality video pairs for training. For long videos, an opening segment sets the style and later windows follow it under a sliding inference scheme, holding motion steady. Across many artistic styles the system performed on par with leading closed source tools in the authors' comparisons, while keeping time consistency that earlier methods lose.

Catalog record

CREATORHuaqiu Li *, Jiahao Wang *, Sijia Cai †, Hualian Sheng, Bing Deng, Jieping Ye, Wenhan Luo †
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.25465

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

EchoStyle: Unlocking High-Fidelity Video Stylization with Reverse Data Synthesis Huaqiu Li *, Jiahao Wang *, Sijia Cai †, Hualian Sheng, Bing Deng, Jieping Ye, Wenhan Luo † Abstract While image stylization has been studied extensively, video stylization remains a critical and largely unsolved challenge in the field of intelligent content creation. Existing methods, usually utilizing a reference image as the style prior, suffer from content leakage, data scarcity and limited adaptability to long videos, leading to suboptimal results with severe style drift and motion distortion. For these issues, we present EchoStyle, a scalable text-driven framework to achieve high-quality stylization of videos with arbitrary lengths. To start with, we construct a video-to-video architecture to appropriately re-fuse the video content and the text style. To address data scarcity, we pioneer an automatic reverse-synthesis pipeline to establish V-Style20k, a large-scale stylization dataset of 20k high-quality video pairs. To facilitate long video stylization, we devise an init-follow-mode mechanism along with a sliding-window inference strategy. Extensive experiments demonstrate EchoStyle’s excellent performance across a wide range of artistic styles, even comparable to leading closed-source solutions. Figure 1 : EchoStyle provides a robust framework for video stylization across expansive artistic domains. By maintaining stringent temporal and motion coherence, it delivers visually compelling results that remain consistent even during long-video inference. 0 0 footnotetext: *: Co-first Author. 1 1 footnotetext: \dagger : Corresponding Author. 1 Introduction Recent advancements in diffusion models have sparked remarkable progress in generative modeling, triggering the evolution of controllable image and video synthesis [ 11 , 36 , 13 , 39 , 43 , 21 , 45 , 50 , 18 , 47 ] . Along with the development, video stylization, which aims to render a given video in a particular style, is garnering growing attention from both academia and industry for its wide-ranging applications such as artistic creation and advertisement production. Previous works in video stylization predominantly focus on introducing a reference image as a style prior. Conventional methods [ 6 , 14 , 10 , 30 ] leverage CNNs or ViTs as backbones, primarily prioritizing real-time efficiency. In recent years, there has been a significant shift towards adopting diffusion models as the foundational architecture. For instance, StyleCrafter [ 23 ] and StyleMaster [ 46 ] adapt pre-trained text-to-video diffusion models [ 2 , 37 ] through post-training strategies. In parallel, training-free methods, such as AnyV2V [ 19 ] and FreeVis [ 42 ] , typically stylize keyframes via image editing models and then propagate the style through feature injection to ensure temporal consistency. Furthermore, frameworks like UniVST [ 32 ] have extended training-free capabilities to support localized video stylization. However, these existing paradigms suffer from several critical limitations: Content-style entanglement. Since the reference image contains redundant style-irrelevant information besides style, the reliance on it often leads to unintended content leakage into the stylized videos [ 20 , 44 ] . Consequently, commercial closed-source video models [ 8 , 35 ] and several image stylization methods [ 20 , 25 , 33 ] concentrate on the text-driven roadmap to achieve concept-level stylization and inherently avoid content leakage. Yet, despite their advancements, a reliable open-source text-driven video stylization paradigm remains elusive. Video data scarcity. The training of video stylization requires high-quality video pairs with the same content but different styles, which are extremely scarce. Existing video stylization methods either bypass training by handcrafted components [ 19 , 42 , 32 ] or conduct training on image datasets [ 3 , 23 , 46 ] , both of which lead to severe style degradation and instability, especially under strong video dynamics. Limited ability of extension. Previous works are confined to short sequences ( \leq 5s) and fail to provide robust, scalable designs for long-duration generation. Therefore, developing a stable and expandable framework for text-driven video stylization remains a valuable and challenging problem. To address the aforementioned challenges, we present EchoStyle, a text-driven framework designed for high-fidelity stylization of videos with arbitrary lengths. We formalize the text-driven video stylization as video-to-video generation and propose a streamlined yet effective architecture based on the Wan2.2-I2V [ 37 ] foundation to appropriately re-fuse the video content and the text style. To bridge the data gap caused by scarcity and inadequate quality, we devise an efficient and reusable data generation pipeline that employs reverse-synthesis to derive reference input videos from open-source stylish datasets. With such a curated pipeline, we construct V-Style20k, a pioneering large-scale video stylization dataset of 20k high-quality video pairs to drive the training process. To further achieve consistent stylization for long videos, we design a temporal recurrent generation approach, which highlights an init-follow-mode mechanism during training and a sliding-window strategy during inference. As shown in Fig. 1 , extensive experiments demonstrate that our method achieves robust and stable stylization for both short and long videos, even in cases of complex content and intense dynamics. We summarize our key contributions as follows: • We propose a novel video-to-video framework tailored for text-driven stylization of videos with arbitrary lengths, which is naturally scalable owing to its simplicity. • We pioneer a fully automated data curation pipeline to construct an innovative large-scale video stylization dataset, V-Style20k, which contains 20k high-quality video pairs. • We incorporate an intricate init-follow-mode mechanism into the training along with a sliding-window inference strategy, enabling stable stylization for long videos spanning several minutes. • Comprehensive experiments demonstrate that EchoStyle achieves superior dynamic style consistency as well as fine-grained content preservation in video stylization, yielding performance comparable to closed-source models. 2 Related Work 2.1 Image & Video Stylization Stylization aims at applying the artistic style from the reference condition to either a given content image or video. For image stylization, early research, such as [ 14 , 6 ] , utilizes a pre-trained network as a style encoder and a carefully designed injection module. More recently, the advance of powerful text-to-image foundation models, such as Stable Diffusion [ 29 ] and FLUX [ 22 ] , has driven a series of methods [ 44 , 28 ] based on diffusion models. Building upon this, works like InstantStyle [ 38 ] , StyleShot [ 7 ] , and B-LoRA [ 5 ] have proposed more granular, time-aware and layer-aware injection strategies to better disentangle the infusion of style and content features. In the field of video stylization, conventional methods [ 6 , 14 , 10 ] predominantly rely on CNN or ViT backbones, prioritizing low-latency inference and real-time efficiency. However, recent years have witnessed a paradigm shift toward diffusion-based architectures. For instance, StyleCrafter [ 23 ] , StyleMaster [ 46 ] and PickStyle [ 26 ] adapt pre-trained T2V models through specialized post-training strategies, while Telestyle [ 49 ] tries to adapt pre-trained image generation models to the video domain. Concurrently, training-free pipelines have gained attention by leveraging non-end-to-end workflows. Approaches like AnyV2V [ 19 ] and FreeVis [ 42 ] typically stylize representative keyframes via image-editing models and subsequently propagate these stylistic attributes through feature injection to ensure temporal consistency. Furthermore, frameworks such as UniVST [ 32 ] have extended these training-free capabilities to support localized or region-aware video stylization. 2.2 Video-to-Video Frameworks Existing research has extensively investigated Video-to-Video (V2V) translation frameworks. In particular, Video-P2P [ 24

Read the complete text