Research map / Video / Video generation
Real-Time Generation of Streamable Talking Portrait Video with Reference-Guided Deep Compression VAEs
Sicheng Xu, Yu Deng, Shoukang Hu, Yichuan Wang, Yizhong Zhang, Zhan Chen, Jiaolong Yang, Baining Guo
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 reference-image guidance (1 ref) to VAE decoder (ablation) | causal video VAE · VoxCeleb2 | PSNR: 29.071 to 31.676 (+2.605 dB) | Improved |
| add reference-image guidance (1 ref) to VAE decoder (ablation) | causal video VAE · HDTF | PSNR: 28.306 to 32.068 (+3.762 dB) | Improved |
| increase reference images in VAE from M=0 to M=3 with CR-VA (ablation) | causal video VAE · VoxCeleb2 | PSNR: 4.375 dB with CR-VA, higher than 3.695 dB without it | Improved |
| increase reference images in VAE from M=0 to M=3 with CR-VA (ablation) | causal video VAE · HDTF | PSNR: 6.696 dB versus 4.843 dB | Improved |
| add causal residual video auto-encoding (CR-VA) without reference guidance (ablation) | causal video VAE · VoxCeleb2 (M=0) | PSNR: 29.071 -> 29.604 | Improved |
| add causal residual video auto-encoding (CR-VA) with 3 references (ablation) | causal video VAE · HDTF (M=3) | PSNR: 33.149 -> 35.374 | Improved |
| increase reference images at generation from M=1 to M=3 | talking portrait generator · HDTF | FVD25: 62.300 -> 43.270 | Improved |
| increase reference images at generation from M=1 to M=3 | talking portrait generator · PortraitOneMin | FVD25: 91.964 -> 73.693 | Improved |
| add classifier-free guidance (la=2, lr=2) vs none (ablation) | talking portrait generator · HDTF (M=1) | FVD25: 101.112 -> 62.300 | Improved |
| raise audio guidance scale la from 2 to 3 (lr=2) (ablation) | talking portrait generator · HDTF (M=1) | FVD25: 62.300 -> 71.464 | Hurt |
| raise audio guidance scale la from 2 to 3 (lr=2) (ablation) | talking portrait generator · HDTF (M=1) | SyncNet confidence S_C: 8.943 -> 9.262 | Improved |
| raise reference guidance scale lr from 2 to 3 (la=2) (ablation) | talking portrait generator · HDTF (M=1) | FVD25: 62.300 -> 80.183 | Hurt |
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.
- Sonic — Sonic: Shifting Focus to Global Audio Perception in Portrait Animation (CVPR 2025) · better on 6 of 6 shared results
- Hallo3 — Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer (CVPR 2025) · better on 2 of 2 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
- Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer (CVPR 2025 · similarity 0.88)
- DAWN: Dynamic Frame Avatar with Non-autoregressive Diffusion Framework for Talking head Video Generation (ICLR 2025 · similarity 0.88)
- Hallo2: Long-Duration and High-Resolution Audio-Driven Portrait Image Animation (ICLR 2025 · similarity 0.87)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.87)
- RealPortrait: Realistic Portrait Animation with Diffusion Transformers (AAAI 2025 · similarity 0.87)
- OmniTalker: One-shot Real-time Text-Driven Talking Audio-Video Generation With Multimodal Style Mimicking (NeurIPS 2025 · similarity 0.86)
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.
