MotionShift: Environment-Compatible Motion Transfer via Planner-Evaluator Synergistic Adaptation
In this work, we propose MotionShift, a novel motion transfer framework that achieves high-fidelity motion transfer while ensuring physical compatibility with the target scene.
ProofPaper ↗
Key points
- Existing motion transfer methods excel at exact, rigid motion replication from reference videos to target scenes, whereas they often fail to account for the compatibility between the transferred motion and the target scene environment in accordance with physical laws.
- In particular, we design a planner--evaluator iterative motion adaptation mechanism, in which a Large Language Model (LLM) acts as the planner to generate motion trajectories by leveraging its strong semantic reasoning capabilities, while a Vision-Language Model (VLM) serves as the evaluator to rectify the motion based on its superior visual perception and scene understanding.
- The adapted motion is then used as a generative condition for an image-to-video (I2V) model to render videos in the target scene.
- Moreover, we show that MotionShift can be readily integrated into existing trajectory-conditioned I2V models, endowing them with realistic motion transfer capabilities.
Sources (1)
- [1]MotionShift: Environment-Compatible Motion Transfer via Planner-Evaluator Synergistic AdaptationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 02:11 PM
In this work, we propose MotionShift, a novel motion transfer framework that achieves high-fidelity motion transfer while ensuring physical compatibility with the target scene.
Existing motion transfer methods excel at exact, rigid motion replication from reference videos to target scenes, whereas they often fail to account for the compatibility between the transferred motion and the target scene environment in accordance with physical laws.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026FlashMap: Flow Map Speculative Decoding
- Oct 16, 2026Deep LLM layers expose passage-level context in EEG during silent reading
- Oct 16, 2026DLite: History-free Speculative Decoding for Agentic AI
- Oct 16, 2026Self-Evolving Coding Agents via Harness-Policy Optimization
- Oct 15, 2026CoLU: Confidence-Weighted Layerwise Unlearning for Large Language Models
- Oct 15, 2026AlphaBet: Reinforcement Learning for LLM Event Forecasting Agents