用於長篇多鏡頭影片生成的多重網格後訓練

摘要

生成長篇多鏡頭影片需要鏡頭內運動的連貫性，以及跨鏡頭視覺一致的敘事。現有影片生成器偏好連續運動，且當整段敘事被壓縮於單一時間軸時，難以呈現完整的鏡頭集合。我們提出 MovieGrid，一種多網格後訓練範式，將長影片分解為較短、依時間排序的區塊，並將其排列於空間網格上以進行聯合建模。此設計減少了每條時間軸所需處理的鏡頭數量，同時促成跨區塊的全域資訊交換。我們從 1,000 部長篇影片建構 Multi-Grid Long Video (MGLV) 資料集，流程包括來源影片收集、階層式分割、網格影片建構，以及角色感知故事標註，產出 54K 部網格影片並搭配故事提示。我們的 Noise-Free Random-Grid Training 保留隨機子集的區塊作為乾淨視覺上下文，用以對其餘區塊去噪。Grid Embedding 編碼網格結構，角色感知 Story Prompts 連結重複出現的實體，而 Grid Boundary Loss 穩定佈局。在相同 token 預算下，MovieGrid 在 1,616 幀影片中生成的鏡頭數是 Temporal Packing 的 6.05 倍。在涵蓋五個真實世界類別的基準上，它達到最先進的鏡頭內一致性（0.9131，相較 HoloCine 的 0.8086）與鏡頭間一致性（0.5914，相較 StoryMem 的 0.5384）。MovieGrid 可透過單次或多次生成，以最小犧牲進一步擴展影片長度。

English

Generating long-form multi-shot videos requires coherent within-shot motion and visually consistent narratives across shots. Existing video generators favor continuous motion and struggle to present complete shot sets when an entire narrative is packed along one temporal axis. We propose MovieGrid, a Multi-Grid Post-Training paradigm that decomposes a long video into shorter, temporally ordered chunks and arranges them on a spatial grid for joint modeling. This design reduces the number of shots handled by each temporal axis while enabling global information exchange across chunks. We construct the Multi-Grid Long Video (MGLV) dataset from 1,000 long-form videos using source video collection, hierarchical segmentation, grid video construction, and character-aware story annotation, producing 54K grid videos paired with story prompts. Our Noise-Free Random-Grid Training retains a random subset of chunks as clean visual context for denoising the remaining chunks. Grid Embedding encodes grid structure, character-aware Story Prompts link recurring entities, and Grid Boundary Loss stabilizes layouts. Under the same token budget, MovieGrid generates 6.05 times more shots than Temporal Packing in a 1,616-frame video. On a benchmark spanning five real-world categories, it achieves state-of-the-art intra-shot consistency (0.9131 versus 0.8086 for HoloCine) and inter-shot consistency (0.5914 versus 0.5384 for StoryMem). MovieGrid can further scale video length with minimal compromise through single or multiple generations.