I'm a beginner in 3dgs and have encountered the following problems，Why is there such a big difference in the effect between the images rendered using the pandaset dataset and the nvs videos

[kffffffff]

Why is there such a big difference between a single image rendered using the Splat model and the rendering results generated from a new NVS viewpoint using the camera_pose.json from Nerfstudio? Could it be because the default time or other parameters need to be adjusted?

nvs:
ns-render camera-path --load-config outputs/pandaset/splatad/2025-11-26_175827/config.yml --camera-path-filename /data/hsk/pandaset/camera_paths/2025-12-01-13-24-34.json --output-path renders/pandaset/2025-12-01-13-24-34.mp4

rander image:

please healp , thx

[kffffffff]