Final-5B — every evaluated result

NVS — re10k128_50f / ncf2, 128 scenes, 48 scored views

Pose — ATE / RPE-rot (lower is better)

Depth — AbsRel / δ<1.25 (lads row)

Text to 4D caption in, RGB + depth + geometry out

Generated from text alone — no RGB frame, no rays, no camera. The caption that produced each clip is burned into the frame so it can be checked against the result directly. Standard classifier-free guidance (scale 6.0), 40 sampling steps, 50 frames.

Point-cloud visualizations RGB + predicted depth, parallax fly-around

Each clip back-projects the input RGB through the model's predicted depth at the predicted camera, then sways a virtual camera around that viewpoint so parallax reveals the 3D structure. Ground truth is shown beside our prediction wherever the benchmark has depth GT. Four scenes per dataset, chosen by our best metrics.

Pose trajectories — 122 sequences (Sim(3) aligned)

scannetv2__scene0707_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0708_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0709_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0710_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0711_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0712_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0713_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0714_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0715_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0716_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0717_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0718_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0719_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0720_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0721_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0722_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0723_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0724_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0725_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0726_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0727_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0728_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0729_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0730_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0731_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0732_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0733_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0734_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0735_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0736_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0737_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0738_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0739_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0740_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0741_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0742_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0743_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0744_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0745_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0746_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0747_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0748_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0749_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0750_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0751_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0752_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0753_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0754_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0755_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0756_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0757_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0758_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0759_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0760_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0761_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0762_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0763_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0764_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0765_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0766_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0767_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0768_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0769_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0770_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0771_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0772_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0773_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0774_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0775_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0776_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0777_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0778_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0779_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0780_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0781_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0782_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0783_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0784_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0785_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0786_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0787_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0788_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0789_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0790_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0791_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0792_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0793_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0794_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0795_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0796_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0797_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0798_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0799_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0800_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0801_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0802_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0803_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0804_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0805_00input clip + Sim(3)-aligned trajectories

scannetv2__scene0806_00input clip + Sim(3)-aligned trajectories

sintel__alley_2input clip + Sim(3)-aligned trajectories

sintel__ambush_4input clip + Sim(3)-aligned trajectories

sintel__ambush_5input clip + Sim(3)-aligned trajectories

sintel__ambush_6input clip + Sim(3)-aligned trajectories

sintel__cave_2input clip + Sim(3)-aligned trajectories

sintel__cave_4input clip + Sim(3)-aligned trajectories

sintel__market_2input clip + Sim(3)-aligned trajectories

sintel__market_5input clip + Sim(3)-aligned trajectories

sintel__market_6input clip + Sim(3)-aligned trajectories

sintel__shaman_3input clip + Sim(3)-aligned trajectories

sintel__sleeping_1input clip + Sim(3)-aligned trajectories

sintel__sleeping_2input clip + Sim(3)-aligned trajectories

sintel__temple_2input clip + Sim(3)-aligned trajectories

sintel__temple_3input clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_sitting_halfsphereinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_sitting_rpyinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_sitting_staticinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_sitting_xyzinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_walking_halfsphereinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_walking_rpyinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_walking_staticinput clip + Sim(3)-aligned trajectories

tum__rgbd_dataset_freiburg3_walking_xyzinput clip + Sim(3)-aligned trajectories