video scanning offloaded to Modal

app.py

@@ -4,9 +4,12 @@ from detector.video import verify_faces_in_video
 from reports.pdf_report import generate_pdf_report
 from utils.youtube_utils import download_youtube_video
 import modal
+
 verify_faces_remote = modal.Function.lookup("deepface-agent", "verify_faces_remote")
 verify_voices_remote = modal.Function.lookup("deepface-agent", "verify_voices_remote")
-
+verify_faces_in_video_remote = modal.Function.lookup(
+    "deepface-agent", "verify_faces_in_video_remote"
+)
 
 
 # Holds latest results
@@ -14,6 +17,7 @@ last_face_result = None
 last_voice_result = None
 last_video_results = None
 
+
 def compare_faces(img1_path: str, img2_path: str) -> str:
     """Use this tool to compare to faces for a match
     Args:
@@ -21,12 +25,12 @@ def compare_faces(img1_path: str, img2_path: str) -> str:
         img2_path: The path to the second image
     """
     global last_face_result
-    
+
     # Read image files as bytes
     with open(img1_path, "rb") as f1, open(img2_path, "rb") as f2:
         img1_bytes = f1.read()
         img2_bytes = f2.read()
-        
+
     result = verify_faces_remote.remote(img1_bytes, img2_bytes)
     result_text = ""
 
@@ -51,12 +55,12 @@ def compare_voices(audio1: str, audio2: str) -> str:
         audio2: The path to the second audio file
     """
     global last_voice_result
-    
+
     try:
         with open(audio1, "rb") as a1, open(audio2, "rb") as a2:
             audio1_bytes = a1.read()
             audio2_bytes = a2.read()
-    
+
         result = verify_voices_remote.remote(audio1_bytes, audio2_bytes)
         result_text = ""
 
@@ -90,7 +94,15 @@ def scan_video(video_file: str, ref_img: str, youtube_url="") -> str:
         except Exception as e:
             return f"❌ Error downloading YouTube video: {str(e)}"
 
-    results = verify_faces_in_video(video_file, ref_img)
+    with open(video_file, "rb") as vf, open(ref_img, "rb") as rf:
+        video_bytes = vf.read()
+        ref_img_bytes = rf.read()
+    try:
+        results = verify_faces_in_video_remote.remote(video_bytes, ref_img_bytes, interval=30)
+        
+    except Exception as e:
+        return f"❌ Error processing  video: {str(e)}"
+
     report = ""
     last_video_results = results
     for r in results:

modal_app/modal_app.py

@@ -77,3 +77,81 @@ def verify_voices_remote(audio1_bytes, audio2_bytes):
     finally:
         pathlib.Path(audio1_path).unlink()
         pathlib.Path(audio2_path).unlink()
+
+
+@app.function(image=image, gpu="any", timeout=600)
+def verify_faces_in_video_remote(video_bytes: bytes, ref_img_bytes: bytes, interval: int = 30):
+    import cv2
+    import tempfile
+    import os
+    import numpy as np
+    from PIL import Image
+    from deepface import DeepFace
+    from io import BytesIO
+
+    results = []
+    frame_paths = []
+
+    with (
+        tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as video_temp,
+        tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as ref_img_temp,
+    ):
+        video_temp.write(video_bytes)
+        ref_img_temp.write(ref_img_bytes)
+        video_path = video_temp.name
+        ref_img_path = ref_img_temp.name
+
+    try:
+        # --- FRAME EXTRACTION ---
+        cap = cv2.VideoCapture(video_path)
+        frame_count = 0
+        temp_dir = tempfile.gettempdir()
+        frame_paths = []
+
+        while cap.isOpened():
+            ret, frame = cap.read()
+            if not ret:
+                break
+
+            if frame_count % interval == 0:
+                frame_path = os.path.join(temp_dir, f"frame_{frame_count}.jpg")
+                success = cv2.imwrite(frame_path, frame)
+                if success:
+                    frame_paths.append((frame_count, frame_path))
+
+            frame_count += 1
+
+        cap.release()
+
+        # --- FACE VERIFICATION ---
+        for frame_id, frame_path in frame_paths:
+            try:
+                result = DeepFace.verify(
+                    img1_path=ref_img_path,
+                    img2_path=frame_path,
+                    enforce_detection=False,
+                )
+                results.append({
+                    "frame": frame_id,
+                    "distance": round(result["distance"], 4),
+                    "verified": result["verified"]
+                })
+            except Exception as e:
+                results.append({
+                    "frame": frame_id,
+                    "error": str(e)
+                })
+
+        return results
+
+    except Exception as e:
+        return [{"error": str(e)}]
+
+    finally:
+        os.remove(video_path)
+        os.remove(ref_img_path)
+        for _, frame_path in frame_paths:
+            try:
+                os.remove(frame_path)
+            except Exception:
+                pass