Research map / Video / Video generation
Gloria: Consistent Character Video Generation via Content Anchors
Yuhang Yang, Fan Zhang, Huaijin Pi, Ailing Zeng, Shuai Guo, Guowei Xu, Wei Zhai, Yang Cao, Zheng-Jun Zha
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add global anchor (ablation) | Gloria · Long-term consistency | ArcFace identity: 0.233 -> 0.787 | Improved |
| add global anchor (ablation) | Gloria · Long-term consistency | subject consistency: 0.840 -> 0.960 | Improved |
| add Superset Content Anchoring (ablation) | Gloria · Long-term consistency | ArcFace identity: 0.489 -> 0.787 | Improved |
| add Superset Content Anchoring (ablation) | Gloria · Long-term consistency | background consistency: 0.899 -> 0.951 | Improved |
| add viewpoint and expression anchors (ablation) | Gloria · Multi-view appearance and expressive ID | Expression ArcFace: 0.668 -> 0.717 | Improved |
| add viewpoint and expression anchors (ablation) | Gloria · Multi-view appearance and expressive ID | DINO-I: 0.801 -> 0.821 | Improved |
| add RoPE as Weak Condition (ablation) | Gloria · Multi-view appearance and expressive ID | Expression ArcFace: 0.671 -> 0.717 | Improved |
| add RoPE as Weak Condition (ablation) | Gloria · Multi-view appearance and expressive ID | DINO-I: 0.817 -> 0.821 | Improved |
| scale model from 5B to 17B (ablation) | Gloria · Model size (10M clips) | ArcFace: 0.583 -> 0.787 | Improved |
| scale training data from 2M to 10M clips (ablation) | Gloria 17B · Data scale | ArcFace: 0.623 -> 0.787 | Improved |
| scale training data from 2M to 10M clips (ablation) | Gloria 17B · Data scale | AES: 3.57 -> 3.63 | Improved |
| add Gemini judge to EmotiEffLib for expression anchors (ablation) | data pipeline · Expression anchor extraction (100 samples) | accuracy: 66% -> 82% | Improved |
Closest papers by meaning
- Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation (CVPR 2024 · similarity 0.85)
- EntityBench: Towards Entity-Consistent Long-Range Multi-Shot Video Generation (NeurIPS 2026 · similarity 0.85)
- Storybooth: Training-Free Multi-Subject Consistency for Improved Visual Storytelling (ICLR 2025 · similarity 0.85)
- Storynizor: Consistent Story Generation via Inter-Frame Synchronized and Shuffled ID Injection (AAAI 2025 · similarity 0.84)
- Advancing Narrative Long Video Generation via Training-Free Identity-Aware Memory (NeurIPS 2026 · similarity 0.84)
- MIMO: Controllable Character Video Synthesis with Spatial Decomposed Modeling (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
