IDK

.gitignore

@@ -102,4 +102,4 @@ ENV/
 *.swp
 *.swo
 
-mcdata/
+test/

main.py

@@ -1,15 +1,18 @@
 import glob
 import os
 import hashlib
-import time
 from PIL import Image, UnidentifiedImageError
 
+import acoustid
+import chromaprint
 from imagehash import average_hash
 from videohash import VideoHash
 from videohash.exceptions import FFmpegFailedToExtractFrames
 
 hashes = set()
+audio_hashes = []
 delete = set()
+hash_lookup = {}
 
 filenames = [x for x in glob.glob('**', recursive=True) if os.path.isfile(x)]
 
@@ -21,8 +24,8 @@ except KeyboardInterrupt:
     print('\nCancelled.')
     os._exit(0)
 
-print('Sorting file list by smallest size...')
+print('Sorting file list by size...')
-filenames.sort(key=os.path.getsize, reverse=False)
+filenames.sort(key=os.path.getsize, reverse=True)
 
 print('Deduplicating by md5 hash...')
 
@@ -50,6 +53,14 @@ def get_image_hash(filename):
     except UnidentifiedImageError:
         return None
 
+def get_audio_hash(filename):
+    try:
+        fp = acoustid.fingerprint_file(filename)[1]
+        values, _ = chromaprint.decode_fingerprint(fp)
+        return chromaprint.hash_fingerprint(values)
+    except acoustid.FingerprintGenerationError:
+        return None
+
 def get_video_hash(filename):
     try:
         return str(VideoHash(path=filename).hash_hex) + '_video'
@@ -66,27 +77,48 @@ for filename in filenames:
     if filename in delete: continue
 
     try:
-        digest = get_image_hash(filename) or get_video_hash(filename)
+        digest = get_image_hash(filename) or get_audio_hash(filename) or get_video_hash(filename)
     except KeyboardInterrupt:
-        print('\nCancelled.')
+        print('Skipping media hashing.')
-        os._exit(0)
+        break
     except BaseException as e:
         print()
         print('Exception', e.__class__.__name__, str(e), 'while hashing:')
         print(filename)
         continue
 
-    time.sleep(5)
-
     if not digest: continue
 
+    hash_lookup[digest] = filename
+
+    if type(digest) == int:
+        for h in audio_hashes:
+            if bin(digest ^ h).count('1') <= 5:   # TODO adjust?
+                delete.add(filename)
+                print()
+                print(digest, filename, 'close to', h, hash_lookup[h])
+                break
+        else:  # for
+            audio_hashes.append(digest)
+    else:
         if digest in hashes:
             delete.add(filename)
         else:
             hashes.add(digest)
 
 print()
-print('Found', len(delete), 'total duplicate files.')
+print()
+for dupe in delete:
+    print(dupe)
+
+print()
+print('Found', len(delete), 'total duplicate files. Delete them?')
+print('ENTER to continue, ctrl+c to cancel.')
+try:
+    input()
+except KeyboardInterrupt:
+    print('\nCancelled.')
+    os._exit(0)
 
 print('Deleting...')
 for dupe in delete:

requirements.txt

@@ -0,0 +1,19 @@
+audioread==3.0.0
+Brotli==1.0.9
+certifi==2022.12.7
+charset-normalizer==3.1.0
+idna==3.4
+imagedominantcolor==1.0.1
+ImageHash==4.3.1
+mutagen==1.46.0
+numpy==1.24.2
+Pillow==9.5.0
+pyacoustid==1.2.2
+pycryptodomex==3.17
+PyWavelets==1.4.1
+requests==2.28.2
+scipy==1.10.1
+urllib3==1.26.15
+videohash==3.0.1
+websockets==11.0
+yt-dlp==2023.3.4