Research map / Video / Video generation
Promptus: Can Prompt Streaming Replace Video Streaming
Jiangkai Wu, Liming Liu, Yunpeng Tan, Junlin Hao, Liang Zhang, Xinggong Zhang
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 |
|---|---|---|---|
| use Promptus instead of H.265 at 140 kbps | Promptus · Mean LPIPS | LPIPS: 0.407 -> 0.289 | Improved |
| use Promptus instead of VAE at 140 kbps | Promptus · Mean LPIPS | LPIPS: 0.428 -> 0.289 | Improved |
| use Promptus instead of VAE at 540 kbps | Promptus · Mean LPIPS | LPIPS: 0.272 -> 0.255 | Improved |
| use Promptus instead of H.266 at 140 kbps | Promptus · Mean LPIPS | LPIPS: 0.332 -> 0.289 | Improved |
| use noisy previous frame instead of random noise as SD input (ablation) | Promptus · Prompt fitting | visual fidelity | Improved |
| combine MSE and LPIPS loss instead of MSE only (ablation) | Promptus · Prompt fitting | visual sharpness | Improved |
| raise prompt rank from 4 to 16 (ablation) | Promptus · Prompt fitting | color fidelity | Improved |
| use 8-bit fitting-aware quantization instead of 32-bit (ablation) | Promptus · Prompt bitrate | quality | No clear effect |
| add prompt regularization toward CLIP embedding mean (ablation) | Promptus · Prompt fitting | fitting failures | Improved |
| use Promptus on detail-rich UVG instead of Animerun relative to H.265 | Promptus · UVG vs Animerun | LPIPS gap: 0.015 -> 0.121 | Improved |
| initialize next frame fitting from previous frame prompt | Promptus · Inversion iterations | iterations to converge: 2000 -> 120 | Improved |
Closest papers by meaning
- Looking Backward: Streaming Video-to-Video Translation with Feature Banks (ICLR 2025 · similarity 0.84)
- Optical-Flow Guided Prompt Optimization for Coherent Video Generation (CVPR 2025 · similarity 0.83)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.83)
- DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation (CVPR 2025 · similarity 0.82)
- Attend Locally, Remember Linearly: Linear Attention as Cross-Frame Memory for Autoregressive Video Diffusion (NeurIPS 2026 · similarity 0.82)
- DeNC: Unleash Neural Codecs in Video Streaming with Diffusion Enhancement (AAAI 2025 · similarity 0.81)
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.
