Gloria：基于内容锚帧的一致角色视频生成
Gloria: Consistent Character Video Generation via Content Anchors
把角色外观压缩成少量「锚帧」作为参考条件，并用 RoPE 位置偏移区分多张参考图、用超集内容锚定抑制复制粘贴，这套「多参考条件去冲突」思路可直接迁移到任何以参考图驱动的视频/图像生成任务。
原文依据 we propose represent the character's visual attributes through a compact set of anchor frames
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论数字角色是现代媒体的核心，但生成具有长时长、多视角外观一致且身份表现力强的角色视频仍然具有挑战性。现有方法要么提供的上下文不足以保持身份，要么利用非角色中心的信息作为「记忆」，导致一致性欠佳。认识到角色视频生成本质上类似于「由外向内看」的场景，本文提出通过一小组紧凑的锚帧来表示角色的视觉属性。该设计为一致性提供了稳定的参考，而基于参考的视频生成本身面临复制粘贴与多参考冲突的挑战。为解决这些问题，我们引入两种机制：超集内容锚定（Superset Content Anchoring），利用训练片段内部与额外的线索来防止重复；以及作为弱条件的 RoPE（RoPE as Weak Condition），通过编码位置偏移来区分多个锚帧。此外，我们构建了一条可扩展的流程，从海量视频中提取这些锚帧。实验表明，我们的方法能够生成超过 10 分钟的高质量角色视频，并在跨视角下实现富有表现力的身份与外观一致性，优于现有方法。
英文原文摘要 1246 字符
Digital characters are central to modern media, yet generating character videos with long-duration, consistent multi-view appearance and expressive identity remains challenging. Existing approaches either provide insufficient context to preserve identity or leverage non-character-centric information as the "memory", leading to suboptimal consistency.Recognizing that character video generation inherently resembles an "outside-looking-in" scenario. In this work, we propose represent the character's visual attributes through a compact set of anchor frames.This design provides stable references for consistency, while reference-based video generation inherently faces challenges of copy-pasting and multi-reference conflicts. To address these, we introduce two mechanisms: Superset Content Anchoring, providing intra- and extra-training clip cues to prevent duplication, and RoPE as Weak Condition, encoding positional offsets to distinguish multiple anchors.Furthermore, we construct a scalable pipeline to extract these anchors from massive videos. Experiments show our method generates high-quality character videos exceeding 10 minutes, and achieves expressive identity and appearance consistency across views, surpassing existing methods.