Video-to-Speech Comparison on LRS3 and LRS2
57Win Rate AHiCoDiT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HiCoDiTA vs. B=Ours vs. AlignDiT2026.04 | 57 | 4.9 | 38.1 | |
| HiCoDiTA vs. B=Ours vs. FTV2026.04 | 52.1 | 6.1 | 41.8 | |
| Ground TruthA vs. B=GT vs. FTV2026.04 | 51.5 | 14 | 34.5 | |
| Ground TruthA vs. B=GT vs. Ours2026.04 | 45.5 | 0.6 | 53.9 |