SAMA: Factorized Semantic Anchoring and Motion Alignment for Instruction-Guided Video Editing

Forum topic · 小凯 · 2026-03-22

Summary

SAMA is a new framework for instruction-guided video editing that factorizes the task into two components: semantic anchoring, which preserves the identity and semantics of source video content, and motion modeling/alignment, which transfers motion faithfully according to the edit instruction. According to the authors (Xinyao Zhang, Wenkai Dong, Yuxin Song; arXiv:2503.16897), SAMA achieves state-of-the-art performance among open-source video editing models and is competitive with leading commercial systems such as Kling-Omni. The paper falls in the computer vision domain and was shared on zhichai.net on 2026-03-19. This page provides the original English abstract alongside a Chinese summary from the forum post, plus a link to the arXiv entry for full details.

Paper Overview

Field: Computer Vision (CV)
Authors: Xinyao Zhang, Wenkai Dong, Yuxin Song
Published: 2026-03-19
arXiv: 2503.16897

Original Abstract

We present SAMA, a framework that factorizes video editing into semantic anchoring and motion modeling. SAMA achieves state-of-the-art performance among open-source models and is competitive with leading commercial systems like Kling-Omni.

Chinese Summary (translated)

The authors propose SAMA, a framework that decomposes video editing into semantic anchoring and motion modeling. SAMA reaches state-of-the-art performance among open-source models and is competitive with leading commercial systems such as Kling-Omni.

Key Points

Factorized design: SAMA splits instruction-guided video editing into two subproblems — semantic anchoring and motion modeling/alignment.
Open-source SOTA: Reported to achieve state-of-the-art results among open-source video editing models.
Commercial-level quality: Competitive with leading commercial systems, including Kling-Omni.

For full technical details, see the arXiv page.