diff --git a/.gitignore b/.gitignore
index cbcbae1..fdb76a2 100644
--- a/.gitignore
+++ b/.gitignore
@@ -7,7 +7,8 @@ build/
 *.so
 .eggs/
 demo_render/
+outputs/
 CLAUDE.md
 .claude/
 .agents/
-skyseg.onnx
\ No newline at end of file
+skyseg.onnx
diff --git a/README.md b/README.md
index d73b1c6..fbdeedb 100644
--- a/README.md
+++ b/README.md
@@ -183,6 +183,19 @@ python demo.py --model_path /path/to/checkpoint.pt \
 | `--point_size` | `0.00001` | Point cloud point size |
 | `--downsample_factor` | `10` | Spatial downsampling for point cloud display |
 
+### Saving Predictions for Later Visualization
+
+Use `--save_predictions` to keep a processed reconstruction and reopen it later
+without rerunning model inference.
+
+```bash
+python demo.py --model_path /path/to/checkpoint.pt \
+    --image_folder /path/to/images/ \
+    --save_predictions outputs/scene.pt
+
+python demo.py --load_predictions outputs/scene.pt
+```
+
 ### Without FlashInfer (SDPA fallback)
 
 ```bash
diff --git a/demo.py b/demo.py
index a8c8636..c2d0099 100644
--- a/demo.py
+++ b/demo.py
@@ -227,6 +227,54 @@ def prepare_for_visualization(predictions, images=None):
     return vis_predictions
 
 
+def save_predictions(path, predictions, images, image_folder=None):
+    """Save postprocessed predictions for later visualization."""
+    os.makedirs(os.path.dirname(os.path.abspath(path)), exist_ok=True)
+    torch.save(
+        {
+            "predictions": predictions,
+            "images": images,
+            "image_folder": image_folder,
+        },
+        path,
+    )
+    print(f"Saved predictions to {path}")
+
+
+def load_saved_predictions(path):
+    """Load predictions saved by --save_predictions."""
+    payload = torch.load(path, map_location="cpu", weights_only=False)
+    if not isinstance(payload, dict) or "predictions" not in payload:
+        raise ValueError(f"{path} is not a saved LingBot-Map predictions file")
+    return (
+        payload["predictions"],
+        payload.get("images"),
+        payload.get("image_folder"),
+    )
+
+
+def launch_viewer(args, predictions, images_cpu, resolved_image_folder):
+    """Launch the Viser point-cloud viewer for postprocessed predictions."""
+    try:
+        from lingbot_map.vis import PointCloudViewer
+        viewer = PointCloudViewer(
+            pred_dict=prepare_for_visualization(predictions, images_cpu),
+            port=args.port,
+            vis_threshold=args.conf_threshold,
+            downsample_factor=args.downsample_factor,
+            point_size=args.point_size,
+            mask_sky=args.mask_sky,
+            image_folder=resolved_image_folder,
+            sky_mask_dir=args.sky_mask_dir,
+            sky_mask_visualization_dir=args.sky_mask_visualization_dir,
+        )
+        print(f"3D viewer at http://localhost:{args.port}")
+        viewer.run()
+    except ImportError:
+        print("viser not installed. Install with: pip install lingbot-map[vis]")
+        print(f"Predictions contain keys: {list(predictions.keys())}")
+
+
 # =============================================================================
 # Main
 # =============================================================================
@@ -290,8 +338,20 @@ def main():
                         help="Save sky mask visualizations (original | mask | overlay) to this directory")
     parser.add_argument("--export_preprocessed", type=str, default=None,
                         help="Export stride-sampled, resized/cropped images to this folder")
+    parser.add_argument("--save_predictions", type=str, default=None,
+                        help="Save postprocessed predictions to this .pt file before visualization")
+    parser.add_argument("--load_predictions", type=str, default=None,
+                        help="Load predictions saved with --save_predictions and launch visualization only")
 
     args = parser.parse_args()
+    if args.load_predictions:
+        if args.video_path is not None:
+            raise ValueError("--load_predictions cannot be combined with --video_path")
+        predictions, images_cpu, saved_image_folder = load_saved_predictions(args.load_predictions)
+        resolved_image_folder = args.image_folder or saved_image_folder
+        launch_viewer(args, predictions, images_cpu, resolved_image_folder)
+        return
+
     assert args.image_folder or args.video_path, \
         "Provide --image_folder or --video_path"
 
@@ -397,26 +457,11 @@ def main():
         images_for_post = images
 
     predictions, images_cpu = postprocess(predictions, images_for_post)
+    if args.save_predictions:
+        save_predictions(args.save_predictions, predictions, images_cpu, resolved_image_folder)
 
     # ── Visualize ────────────────────────────────────────────────────────────
-    try:
-        from lingbot_map.vis import PointCloudViewer
-        viewer = PointCloudViewer(
-            pred_dict=prepare_for_visualization(predictions, images_cpu),
-            port=args.port,
-            vis_threshold=args.conf_threshold,
-            downsample_factor=args.downsample_factor,
-            point_size=args.point_size,
-            mask_sky=args.mask_sky,
-            image_folder=resolved_image_folder,
-            sky_mask_dir=args.sky_mask_dir,
-            sky_mask_visualization_dir=args.sky_mask_visualization_dir,
-        )
-        print(f"3D viewer at http://localhost:{args.port}")
-        viewer.run()
-    except ImportError:
-        print("viser not installed. Install with: pip install lingbot-map[vis]")
-        print(f"Predictions contain keys: {list(predictions.keys())}")
+    launch_viewer(args, predictions, images_cpu, resolved_image_folder)
 
 
 if __name__ == "__main__":