Research map / Video / Video generation
STIV: Scalable Text and Image Conditioned Video Generation
Zongyu Lin, Wei Liu, Chen Chen, Jiasen Lu, Wenze Hu, Tsu-Jui Fu, Jesse Allardice, Zhengfeng Lai, Liangchen Song, Bowen Zhang, Cha Chen, Yiran Fei and 4 more
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 flow matching training objective (ablation) | T2I-XL · T2I (GenEval) | GenEval: 0.379 -> 0.457 | Improved |
| add CFG renormalization at inference (ablation) | T2I-XL · COCO | FID: 24.96 -> 21.16 | Improved |
| add internal bigG CLIP text encoder (ablation) | T2I-XL · T2I (GenEval) | GenEval: 0.492 -> 0.607 | Improved |
| add synthetic recaptions to original captions (ablation) | T2I-XL · T2I (GenEval) | GenEval: 0.607 -> 0.685 | Improved |
| add MaskDiT 50% spatial masking (ablation) | T2I-XL · T2I | FID and GenEval: FID 20.26 -> 23.85; GenEval 0.474 -> 0.499 | Mixed |
| use frame replacement instead of cross attention alone (ablation) | STIV-XL · VBench-I2V | Average scores: 73.0 -> 77.3 | Improved |
| add large projection layer to FR + CA (ablation) | STIV-XL · VBench-I2V | Dynamic Degree: 35.4 -> 22.4 | Hurt |
| add first frame loss to FR + CA + LP (ablation) | STIV-XL · VBench-I2V | Average scores: 75.6 -> 75.3 | Hurt |
| add image condition dropout with JIT-CFG to a TI2V model (ablation) | STIV-M-512 · VBench T2V | Total score: 66.1 -> 80.7 | Improved |
| add image condition dropout with JIT-CFG (ablation) | STIV-M-512 · VBench | Dynamic Degree: 10.2 -> 24.0 | Improved |
| add image condition dropout with JIT-CFG (ablation) | STIV-M-512 · VBench-I2V | Total score: 90.1 -> 89.8 | Hurt |
| initialize TI2V from T2V instead of T2I (ablation) | STIV-XL · VBench-I2V | Cam Mot: 29.8 -> 38.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CogVideoX-5B — CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer (ICLR 2025) · better on 2 of 2 shared results
- LaVie-2 — LaVie: High-Quality Video Generation with Cascaded Latent Diffusion Models (IJCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TI2V-Zero: Zero-Shot Image Conditioning for Text-to-Video Diffusion Models (CVPR 2024 · similarity 0.91)
- A Recipe for Scaling up Text-to-Video Generation with Text-free Videos (CVPR 2024 · similarity 0.90)
- Encapsulated Composition of Text-to-Image and Text-to-Video Models for High-Quality Video Synthesis (CVPR 2025 · similarity 0.89)
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.88)
- LinGen: Towards High-Resolution Minute-Length Text-to-Video Generation with Linear Computational Complexity (CVPR 2025 · similarity 0.88)
- Grid Diffusion Models for Text-to-Video Generation (CVPR 2024 · similarity 0.88)
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.
