chenkangjie1123/VGGT-Diff-Trajectory-480x832

VGGT-Diff Trajectory 480x832 (36K)

This checkpoint generates an 80-frame continuous camera trajectory from six posed source views. It also supports pose-free inference by first estimating the six source cameras with VGGT-Omega.

Configuration

• Resolution: 480 x 832
• Task: continuous camera-trajectory video generation
• Input: six posed source views and an ordered target camera path
• Output: an 80-frame novel-view video
• VAE: sources are encoded independently; target video uses causal 4:1 temporal compression
• Geometry: frozen VGGT-Omega features with temporally packed camera conditioning
• Training: full-model trajectory fine-tuning on curated sequential DL3DV data

Pose-free usage

python scripts/infer_pose_free.py \
  --checkpoint /path/to/vggt-diff.safetensors \
  --omega-checkpoint /path/to/vggt_omega_1b_512.pt \
  --source-dir /path/to/six_rgb_images \
  --output outputs/pose_free

VGGT-Omega weights and Wan2.1 base-model files are separate dependencies and are not redistributed in this repository.

Links

• Hugging Face Paper
• arXiv
• Code
• Project page

License

Released under the non-commercial VGGT-Diff Research License. Third-party models and datasets remain subject to their own terms.