|
|
import modal |
|
|
from utils.youtube_utils import download_youtube_video |
|
|
from reports.pdf_report import generate_pdf_report |
|
|
|
|
|
last_face_result = "" |
|
|
last_voice_result = "" |
|
|
last_video_results = [] |
|
|
|
|
|
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" |
|
|
) |
|
|
|
|
|
|
|
|
def compare_faces(img1_path: str, img2_path: str) -> str: |
|
|
"""Use this tool to compare to faces for a match |
|
|
Args: |
|
|
img1_path: The path to the first image |
|
|
img2_path: The path to the second image |
|
|
""" |
|
|
global last_face_result |
|
|
|
|
|
|
|
|
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 = "" |
|
|
|
|
|
if "error" in result: |
|
|
return f"❌ Error: {result['error']}" |
|
|
|
|
|
if result["verified"]: |
|
|
result_text = f"✅ Match! Distance: {result['distance']:.4f} (Threshold: {result['threshold']})" |
|
|
last_face_result = result_text |
|
|
return result_text |
|
|
|
|
|
else: |
|
|
result_text = f"❌ No Match. Distance: {result['distance']:.4f} (Threshold: {result['threshold']})" |
|
|
last_face_result = result_text |
|
|
return result_text |
|
|
|
|
|
|
|
|
def compare_voices(audio1: str, audio2: str) -> str: |
|
|
"""Use this tool to compare two voices for a match |
|
|
Args: |
|
|
audio1: The path to the first audio file |
|
|
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 = "" |
|
|
|
|
|
if "error" in result: |
|
|
return f"❌ Error: {result['error']}" |
|
|
|
|
|
if result["match"]: |
|
|
result_text = f"✅ Same speaker detected. Similarity: {result['similarity']} (Threshold: {result['threshold']})" |
|
|
last_voice_result = result_text |
|
|
return result_text |
|
|
else: |
|
|
result_text = f"❌ Different speakers. Similarity: {result['similarity']} (Threshold: {result['threshold']})" |
|
|
last_voice_result = result_text |
|
|
return result_text |
|
|
except Exception as e: |
|
|
return f"❌ Error reading audio files: {str(e)}" |
|
|
|
|
|
|
|
|
def scan_video(video_file: str, ref_img: str, youtube_url="") -> str: |
|
|
"""Use this tool to scan a video for deepfake face swaps |
|
|
Args: |
|
|
video_file: The path to the video file |
|
|
ref_img: The path to the reference image |
|
|
youtube_url: The YouTube URL (optional) |
|
|
""" |
|
|
global last_video_results |
|
|
|
|
|
if youtube_url: |
|
|
try: |
|
|
video_file = download_youtube_video(youtube_url) |
|
|
except Exception as e: |
|
|
return f"❌ Error downloading YouTube video: {str(e)}" |
|
|
|
|
|
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: |
|
|
if "error" in r: |
|
|
report += f"\n⚠️ Frame {r['frame']}: {r['error']}" |
|
|
|
|
|
else: |
|
|
status = "✅ Match" if r["verified"] else "❌ Mismatch" |
|
|
report += f"\n🖼 Frame {r['frame']}: {status} (Distance: {r['distance']})" |
|
|
|
|
|
return report |
|
|
|
|
|
|
|
|
|
|
|
def generate_report(): |
|
|
return generate_pdf_report(last_face_result, last_voice_result, last_video_results) |
