Research map / Video / Video generation
Phenaki: Variable Length Video Generation from Open Domain Textual Descriptions
Ruben Villegas, Mohammad Babaeizadeh, Pieter-Jan Kindermans, Hernan Moraldo, Han Zhang, Mohammad Taghi Saffar, Santiago Castro, Julius Kunze, Dumitru Erhan
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 |
|---|---|---|---|
| train with more image data rather than video-only data (ablation) | Phenaki · Text-to-video | CLIP: 0.298 -> 0.303 | Improved |
| train with more image data rather than video-only data (ablation) | Phenaki · Text-to-video | FVD: 168.9 -> 198.4 | Hurt |
| train with image data rather than video-only data (ablation) | Phenaki · Text-to-image | FID: 53.9 -> 29.4 | Improved |
| increase image data from 20% to 50% (ablation) | Phenaki · Text-to-video | FVD: 198.4 -> 239.7 | Hurt |
| use C-ViViT temporal-spatial video encoding instead of per-frame ViT VQ-GAN | Video quantization · Moments-in-Time | FVD: 166.6 -> 65.78 | Improved |
| use C-ViViT temporal-spatial video encoding instead of per-frame ViT VQ-GAN | Video quantization · Moments-in-Time | Number of Tokens: 2560 -> 1536 | Improved |
| use C-ViViT instead of per-frame ViT VQ-GAN | Video quantization · Moments-in-Time | FID: 3.4 -> 4.5 | Hurt |
| use Phenaki for video prediction rather than the strongest listed baseline | Video prediction · Kinetics-600 | FVD: 16.2 ± 0.34 -> 36.4 ± 0.19 | Hurt |
| use Phenaki for video prediction rather than the strongest listed baseline | Video prediction · BAIR | FVD: 84.2 -> 97.0 | Hurt |
| use zero-shot Phenaki instead of NUWA | Text-to-video generation · Kinetics-400 | FID Video: 7.05 -> 3.84 | Improved |
Later papers that compared with this method
4 later accepted papers report Phenaki as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- CanvasMAR: Improving Masked Autoregressive Video Prediction With Canvas (NeurIPS 2026 · better on 2 of 2 shared results)
- Improving Autoregressive Video Modeling with History Understanding (ICLR 2026 · better on 1 of 1 shared result)
- AdapTok: Learning Adaptive and Temporally Causal Video Tokenization in a 1D Latent Space (CVPR 2026 · better on 1 of 1 shared result)
- Vlogger: Make Your Dream A Vlog (CVPR 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
- Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization (ICML 2024 · similarity 0.86)
- LARP: Tokenizing Videos with a Learned Autoregressive Generative Prior (ICLR 2025 · similarity 0.86)
- Mimir: Improving Video Diffusion Models for Precise Text Understanding (CVPR 2025 · similarity 0.85)
- VQToken: Neural Discrete Token Representation Learning for Extreme Token Reduction in Video Large Language Models (NeurIPS 2025 · similarity 0.85)
- InstanceCap: Improving Text-to-Video Generation via Instance-aware Structured Caption (CVPR 2025 · similarity 0.85)
- The Devil is in the Prompts: Retrieval-Augmented Prompt Optimization for Text-to-Video Generation (CVPR 2025 · similarity 0.85)
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.
