Point4D chains motion reconstruction across overlapping video chunks by re-querying each predicted 3D endpoint directly in the next chunk — no reprojection, no matching — enabling robust 4D tracking over 300+ frames.
We introduce Point4D, a feed-forward model for 4D reconstruction of long-range video sequences. Point4D is able to reliably infer dense per-point 3D trajectories across multi-hundred video sequences, unlike existing 4D methods that are limited to short input windows of at most a few dozen frames. A key innovation that enables this is our flexible 3D query-based motion decoder that decouples trajectory prediction from image-plane visibility. The predicted 3D endpoints are then directly re-queried in the next chunk without re-projection or matching. Furthermore, we show that extracting and reusing a visual descriptor from an arbitrary frame where the point is visible leads to superior performance than purely relying on the source patch. Overall, Point4D achieves state-of-the-art performance across diverse long-video tracking benchmarks spanning over 200 frames and substantially outperforms previous feed-forward 4D methods.
On ~200-frame sequences (chunk 48 / overlap 8), Point4D outperforms every feed-forward baseline across datasets and visibility conditions.
Iterative trackers (SpaTrackerV2, TAPIP3D) are compared in the paper — Point4D outperforms or matches their accuracy while running an order of magnitude faster. Full tables and additional datasets in the paper.
Per-chunk trajectory error as the video progresses — Point4D's (red plot) error grows slowest across chunk boundaries.

^ Dynamic Replica (200 frames)

^ Panoptic Studio (150 frames)
Per-chunk EPE (↓) and APD (↑); chunk 48 / overlap 8.
Point4D performs comparably within a single chunk (up to 64 frames) — its gains come from chaining.
| Method | LSFOdyssey | Dyn. Replica | PStudio | ||||
|---|---|---|---|---|---|---|---|
| EPE↓ | APD↑ | EPE↓ | APD↑ | EPE↓ | APD↑ | ||
| 2D | DA3 + CoTracker3 | 0.50 | 0.69 | 0.11 | 0.89 | 0.33 | 0.63 |
| Iter. | TAPIP3D | 0.35 | 0.66 | 0.87 | 0.50 | 0.30 | 0.65 |
| SpatialTrackerV2 | 0.34 | 0.68 | 0.69 | 0.62 | 0.21 | 0.75 | |
| FF | St4RTrack | 0.56 | 0.48 | 0.17 | 0.81 | 0.41 | 0.53 |
| Trace Anything | 0.76 | 0.42 | 0.34 | 0.66 | 0.25 | 0.72 | |
| Any4D | 0.27 | 0.72 | 0.07 | 0.93 | 0.28 | 0.66 | |
| V-DPM | 0.14 | 0.85 | 0.14 | 0.84 | 0.17 | 0.79 | |
| 4RC | 0.16 | 0.87 | 0.07 | 0.95 | 0.29 | 0.66 | |
| Point4D (Ours) | 0.27 | 0.73 | 0.09 | 0.91 | 0.23 | 0.73 | |
EPE (↓) and APD (↑); 1st 2nd 3rd.
Pick a scene on the left to load it, then drag to rotate and scroll to zoom. The toolbar sets the trail length and toggles the trajectory overlay.