The Year in AI Papers
The year/Z.ai / GLM

Paper 2606.10804

SCAIL-2: Unifying Controlled Character Animation with End-to-end In-Context Conditioning

Wenhao Yan, Fengjia Guo, Zhuoyi Yang, Jie Tang

Published
Jun 2026
Research lab
Z.ai / GLM
Citations
1
GitHub
1.1K stars

01 In brief

Summary

SCAIL-2 is an end-to-end framework for controlled character animation that bypasses intermediate representations like pose skeletons or masked backgrounds, which cause information loss.

It directly concatenates driving videos to the sequence, allowing the model to capture all visual information.

To address the lack of end-to-end data, the authors unify sub-tasks (character image animation and replacement) with decoupled conditions and synthesize MotionPair-60K, a dataset of 59,376 motion-transfer pairs.

The model uses in-context mask conditioning and mode-specific RoPE as soft guidance.

Bias-Aware DPO is introduced to mitigate synthetic data bias, especially in hand regions.

Experiments show SCAIL-2 outperforms state-of-the-art methods in various tasks, including cross-identity motion following, environment integration, and multi-character interactions.

The model is trained on a 14B I2V backbone (Wan2.1-14B-I2V) with 64 H100 GPUs for about a week.

The framework demonstrates gains from end-to-end training and unification under reverse-driving, enabling compositional tasks.

Limitations include dependence on large-scale paired data and generator fidelity.

02 From the paper

Abstract

Controlled character animation requires transferring motion from a driving sequence to a reference character. Prior works heavily rely on intermediate representations, including pose skeletons to represent motion or masked background to represent environment, which inevitably leads to information loss. To address this, we present SCAIL-2, a framework that bypasses those intermediates and achieves \textbf{end-to-end} character animation. By directly concatenating driving videos to the sequence, the model can obtain all the required visual information from the input video. To address the lack of end-to-end data, we unify sub-tasks of character animation with decoupled conditions and then curate a pipeline to synthesize MotionPair-60K, an end-to-end motion transfer dataset containing heterogeneous tasks of character animation. To achieve the unification, we utilize in-context mask conditioning and mode-specific RoPE as soft guidance beyond textual instructions and raw visual information. To address synthetic discrepancy in detailed regions, we propose Bias-Aware DPO to construct preference items to mitigate the errors. Extensive experiments demonstrate that our method substantially outperforms existing state-of-the-art approaches in various character animation tasks. A large subset of synthetic data as well as model weights will be released at our project page: https://teal024.github.io/SCAIL-2/.