Research map / Video / Video generation
VideoStudio: Generating Consistent-Content and Multi-Scene Videos
Fuchen Long, Zhaofan Qiu, Ting Yao, Tao Mei
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 entity reference images across scenes (ablation) | VideoStudio · ActivityNet Captions | scene consistency: 50.8 -> 75.1 | Improved |
| add entity reference images across scenes (ablation) | VideoStudio · Coref-SV | scene consistency: 40.9 -> 77.3 | Improved |
| add entity reference images across scenes (ablation) | VideoStudio · ActivityNet Captions | FVD: 624 -> 395 | Improved |
| add HD-VG-130M subset to training data (ablation) | VideoStudio w/o Ref. · ActivityNet Captions | FVD: 789 -> 624 | Improved |
| add action category guidance to VideoStudio-Vid (ablation) | VideoStudio-Vid · WebVid-10M val (real frame reference) | FVD: 157.3 -> 116.5 | Improved |
| add action category condition (ablation) | VideoStudio-Vid · VBench action score | action score: 90.3% -> 96.5% | Improved |
| add LLM-generated script and reference images to multi-scene generation (ablation) | VideoStudio · Multi-scene user study | content consistency rank: 3.4 -> 1.2 | Improved |
| use LLM script generation instead of duplicated prompt (ablation) | VideoStudio w/o Ref. · Multi-scene user study | logical coherence rank: 3.0 -> 2.3 | Improved |
| use LLM script without reference images (ablation) | VideoStudio w/o Ref. · Multi-scene user study | content consistency rank: 2.3 -> 3.4 | Hurt |
| replace ChatGLM3-6B with GPT-4 for script generation (ablation) | VideoStudio · user preference on videos | preference share: 20% vs 21%, tie 59% | No clear effect |
| replace ChatGLM3-6B with GPT-4 for script generation (ablation) | VideoStudio · user preference on scripts | preference share: 25% -> 37% | Improved |
| use SD frame plus VideoStudio-Vid instead of ModelScopeT2V on MSR-VTT | text-to-video · MSR-VTT val | FID: 11.1 -> 11.9 | Hurt |
Later papers that compared with this method
3 later accepted papers report VideoStudio as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- FilmWeaver: Weaving Consistent Multi-Shot Videos with Cache-Guided Autoregressive Diffusion (AAAI 2026 · better on 5 of 5 shared results)
- Long Context Tuning for Video Generation (ICCV 2025 · better on 3 of 5 shared results)
- Mask^2DiT: Dual Mask-based Diffusion Transformer for Multi-Scene Long Video Generation (CVPR 2025 · better on 4 of 4 shared results)
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
- Vivid-ZOO: Multi-View Video Generation with Diffusion Model (NeurIPS 2024 · similarity 0.90)
- SEINE: Short-to-Long Video Diffusion Model for Generative Transition and Prediction (ICLR 2024 · similarity 0.90)
- LLM-grounded Video Diffusion Models (ICLR 2024 · similarity 0.89)
- MeDM: Mediating Image Diffusion Models for Video-to-Video Translation with Temporal Correspondence Guidance (AAAI 2024 · similarity 0.89)
- EIDT-V: Exploiting Intersections in Diffusion Trajectories for Model-Agnostic, Zero-Shot, Training-Free Text-to-Video Generation (CVPR 2025 · similarity 0.89)
- StoryDiffusion: Consistent Self-Attention for Long-Range Image and Video Generation (NeurIPS 2024 · similarity 0.89)
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.
