diff --git a/tools/realtime_infer.py b/tools/realtime_infer.py
new file mode 100644
index 00000000..34b896a8
--- /dev/null
+++ b/tools/realtime_infer.py
@@ -0,0 +1,109 @@
+#!/usr/bin/env python
+# coding=utf-8
+"""Realtime multi-camera inference script for BEVFormer.
+
+This script loads a BEVFormer model and performs realtime
+inference on multiple camera streams. It accepts camera
+indices or video file paths as input and prints the
+prediction results. Optionally it can display the camera
+feed and save results.
+"""
+
+import argparse
+import os
+import os.path as osp
+import sys
+from typing import List
+
+import cv2
+import numpy as np
+import torch
+from mmcv import Config
+from mmcv.runner import load_checkpoint
+from mmcv.parallel import collate, scatter
+
+FILE_DIR = osp.dirname(osp.abspath(__file__))
+ROOT_DIR = osp.dirname(FILE_DIR)
+sys.path.append(ROOT_DIR)
+
+from projects.mmdet3d_plugin.datasets.pipelines import Compose
+from mmdet3d.models import build_model
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description='Realtime inference with multi-camera input')
+    parser.add_argument('config', help='config file path')
+    parser.add_argument('checkpoint', help='checkpoint file')
+    parser.add_argument(
+        '--cams',
+        default='0',
+        help='camera ids or video paths separated by comma, e.g. "0,1"')
+    parser.add_argument('--device', default='cuda:0', help='device id')
+    parser.add_argument('--save-dir', default=None, help='directory to save images')
+    parser.add_argument('--no-show', action='store_true', help='do not display window')
+    parser.add_argument('--fps', type=float, default=10.0, help='refresh rate')
+    return parser.parse_args()
+
+
+def build_data_pipeline(cfg) -> Compose:
+    pipeline = []
+    for step in cfg.data.test.pipeline:
+        if step['type'] != 'LoadMultiViewImageFromFiles':
+            pipeline.append(step)
+    return Compose(pipeline)
+
+
+def open_cameras(cam_str: str) -> List[cv2.VideoCapture]:
+    cams = []
+    for src in cam_str.split(','):
+        src = src.strip()
+        cap = cv2.VideoCapture(int(src) if src.isdigit() else src)
+        if not cap.isOpened():
+            raise RuntimeError(f'Cannot open camera source: {src}')
+        cams.append(cap)
+    return cams
+
+
+def main():
+    args = parse_args()
+
+    cfg = Config.fromfile(args.config)
+    cfg.model.pretrained = None
+    model = build_model(cfg.model, test_cfg=cfg.get('test_cfg'))
+    load_checkpoint(model, args.checkpoint, map_location='cpu')
+    model.to(args.device)
+    model.eval()
+
+    pipeline = build_data_pipeline(cfg)
+
+    caps = open_cameras(args.cams)
+
+    while True:
+        frames = []
+        for cap in caps:
+            ret, frame = cap.read()
+            if not ret:
+                return
+            frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            frames.append(frame)
+        data = dict(img=frames, lidar2img=[np.eye(4) for _ in frames])
+        data = pipeline(data)
+        data = collate([data], samples_per_gpu=1)
+        data = scatter(data, [args.device])[0]
+        with torch.no_grad():
+            result = model(return_loss=False, rescale=True, **data)
+        print(result)
+
+        if not args.no_show:
+            for idx, frame in enumerate(frames):
+                cv2.imshow(f'cam{idx}', cv2.cvtColor(frame, cv2.COLOR_RGB2BGR))
+            if cv2.waitKey(int(1000 / args.fps)) & 0xFF == ord('q'):
+                break
+    for cap in caps:
+        cap.release()
+    cv2.destroyAllWindows()
+
+
+if __name__ == '__main__':
+    main()