Identity-Preserving Text-to-Video Generation on VIP-200K (test)
0.631FaceSim-ArcST-DRC
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ST-DRCReference Concatenation=true, TASS-RoPE=true, Reference Augmentation=true, Auxiliary ID Loss=true, Decoupled CFG=true2026.06 | 0.631 | 0.671 | 33.04 | 0.576 | 0.682 | 0.992 | 0.93 | |
| LTX-2.3-Base + Aux. ID LossReference Concatenation=true, TASS-RoPE=true, Reference Augmentation=true, Auxiliary ID Loss=true, Decoupled CFG=false2026.06 | 0.618 | 0.658 | 32.28 | 0.568 | 0.68 | 0.986 | 0.88 | |
| LTX-2.3-Base + Ref. Aug.Reference Concatenation=true, TASS-RoPE=true, Reference Augmentation=true, Auxiliary ID Loss=false, Decoupled CFG=false2026.06 | 0.592 | 0.631 | 32.12 | 0.563 | 0.679 | 0.979 | 0.89 | |
| LTX-2.3-Base + TASS-RoPEReference Concatenation=true, TASS-RoPE=true, Reference Augmentation=false, Auxiliary ID Loss=false, Decoupled CFG=false2026.06 | 0.573 | 0.612 | 31.94 | 0.555 | 0.677 | 0.972 | 0.86 | |
| ConsisID2026.06 | 0.566 | 0.593 | 31.47 | 0.526 | 0.67 | 0.982 | 0.51 | |
| LTX-2.3-Base + Ref. Concat.Reference Concatenation=true, TASS-RoPE=false, Reference Augmentation=false, Auxiliary ID Loss=false, Decoupled CFG=false2026.06 | 0.541 | 0.579 | 31.76 | 0.548 | 0.674 | 0.961 | 0.83 | |
| Phantom2026.06 | 0.537 | 0.627 | 31.55 | 0.532 | 0.668 | 0.982 | 0.87 | |
| IPVG-STD2026.06 | 0.492 | 0.519 | 31.82 | 0.561 | 0.683 | 0.986 | 0.88 | |
| VACE2026.06 | 0.423 | 0.444 | 32.61 | 0.564 | 0.688 | 0.986 | 0.71 | |
| LTX-2.3-BaseReference Concatenation=false, TASS-RoPE=false, Reference Augmentation=false, Auxiliary ID Loss=false, Decoupled CFG=false2026.06 | 0.284 | 0.317 | 32.18 | 0.558 | 0.681 | 0.985 | 0.87 |