diff --git a/akarpov/music/services/file.py b/akarpov/music/services/file.py
index 95bc450..054506c 100644
--- a/akarpov/music/services/file.py
+++ b/akarpov/music/services/file.py
@@ -7,11 +7,18 @@
 import librosa
 import mutagen
 import numpy as np
+from django.core.files import File
+from django.db import IntegrityError, transaction
+from django.db.models import Count
+from django.utils.text import slugify
 from mutagen.id3 import ID3
+from mutagen.mp3 import MP3
 from PIL import Image, UnidentifiedImageError
 
-from akarpov.music.models import Song
+from akarpov.music.models import Album, Author, Song
 from akarpov.music.services.db import load_track
+from akarpov.users.models import User
+from akarpov.utils.generators import generate_charset
 
 
 def load_dir(path: str, user_id: int):
@@ -227,3 +234,316 @@ def _process_file(self, path: str, info: dict, user_id: int):
         except Exception as e:
             print("File processing failed", file=path, error=str(e))
             self.failed_files.append(path)
+
+
+def clean_title_for_slug(title: str) -> str:
+    """Clean title for slug generation."""
+    # Remove common suffixes
+    suffixes = [
+        "(Original Mix)",
+        "(Radio Edit)",
+        "(Extended Mix)",
+        "(Official Video)",
+        "(Music Video)",
+        "(Lyric Video)",
+        "(Audio)",
+        "(Official Audio)",
+        "(Visualizer)",
+        "(Official Music Video)",
+        "(Official Lyric Video)",
+    ]
+    cleaned = title
+    for suffix in suffixes:
+        cleaned = cleaned.replace(suffix, "")
+    return cleaned.strip()
+
+
+def process_authors_string(authors_str: str) -> list[str]:
+    """Split author string into individual author names."""
+    if not authors_str:
+        return []
+
+    # First split by major separators
+    authors = []
+    for part in authors_str.split("/"):
+        for subpart in part.split("&"):
+            # Handle various featuring cases
+            for feat_marker in [
+                " feat.",
+                " ft.",
+                " featuring.",
+                " presents ",
+                " pres. ",
+            ]:
+                if feat_marker in subpart.lower():
+                    parts = subpart.lower().split(feat_marker, 1)
+                    authors.extend(part.strip() for part in parts)
+                    break
+            else:
+                # Handle collaboration markers
+                if " x " in subpart:
+                    authors.extend(p.strip() for p in subpart.split(" x "))
+                else:
+                    authors.append(subpart.strip())
+
+    # Remove duplicates while preserving order
+    seen = set()
+    return [x for x in authors if not (x.lower() in seen or seen.add(x.lower()))]
+
+
+def extract_mp3_metadata(file_path: str) -> dict | None:
+    """Extract metadata from MP3 file."""
+    try:
+        audio = MP3(file_path, ID3=ID3)
+        tags = audio.tags if audio.tags else {}
+
+        # Get filename without extension for fallback
+        base_filename = os.path.splitext(os.path.basename(file_path))[0]
+
+        metadata = {
+            "title": None,
+            "album": None,
+            "artists": None,
+            "genre": None,
+            "release_year": None,
+            "length": audio.info.length,
+            "image_data": None,
+            "image_mime": None,
+        }
+
+        if tags:
+            # Extract basic metadata with fallbacks
+            metadata["title"] = str(tags.get("TIT2", "")) or base_filename
+            metadata["album"] = str(tags.get("TALB", ""))
+            metadata["artists"] = str(tags.get("TPE1", "")) or str(tags.get("TPE2", ""))
+            metadata["genre"] = str(tags.get("TCON", ""))
+            metadata["release_year"] = str(tags.get("TDRC", "")) or str(
+                tags.get("TYER", "")
+            )
+
+            # Extract cover art
+            for tag in tags.getall("APIC"):
+                if tag.type == 3:  # Front cover
+                    metadata["image_data"] = tag.data
+                    metadata["image_mime"] = tag.mime
+                    break
+
+            # Clean up title if it came from filename
+            if metadata["title"] == base_filename:
+                parts = base_filename.split(" - ", 1)
+                if len(parts) > 1 and not metadata["artists"]:
+                    metadata["artists"] = parts[0]
+                    metadata["title"] = parts[1]
+
+        return metadata
+    except Exception as e:
+        print(f"Error extracting metadata from {file_path}: {str(e)}")
+        return None
+
+
+def generate_unique_slug(
+    base_name: str, model_class, existing_id=None, max_length=20
+) -> str:
+    """Generate a unique slug for a model instance."""
+    # Clean and slugify the base name
+    slug = slugify(clean_title_for_slug(base_name))
+
+    # Truncate if necessary
+    if len(slug) > max_length:
+        slug = slug[:max_length].rsplit("-", 1)[0]
+
+    original_slug = slug
+    counter = 1
+
+    # Check for uniqueness
+    while True:
+        if existing_id:
+            exists = (
+                model_class.objects.filter(slug=slug).exclude(id=existing_id).exists()
+            )
+        else:
+            exists = model_class.objects.filter(slug=slug).exists()
+
+        if not exists:
+            break
+
+        # Generate new slug
+        if counter == 1:
+            if len(original_slug) > (max_length - 7):  # Leave room for _XXXXX
+                base = original_slug[: (max_length - 7)]
+            else:
+                base = original_slug
+            slug = f"{base}_{generate_charset(5)}"
+        else:
+            if len(original_slug) > (max_length - len(str(counter)) - 1):
+                base = original_slug[: (max_length - len(str(counter)) - 1)]
+            else:
+                base = original_slug
+            slug = f"{base}_{counter}"
+
+        counter += 1
+
+    return slug
+
+
+def save_image_as_png(image_data: bytes, mime_type: str) -> str | None:
+    """Convert image data to PNG and save temporarily."""
+    try:
+        if not image_data:
+            return None
+
+        img = Image.open(BytesIO(image_data))
+        temp_path = f"/tmp/{generate_charset(10)}.png"
+        img.save(temp_path, "PNG")
+        return temp_path
+    except Exception as e:
+        print(f"Error processing image: {str(e)}")
+        return None
+
+
+def get_or_create_album(album_name: str, authors: list[Author]) -> Album | None:
+    """Get or create album with proper locking and uniqueness check."""
+    if not album_name:
+        return None
+
+    with transaction.atomic():
+        # Try to find existing album
+        album = (
+            Album.objects.select_for_update().filter(name__iexact=album_name).first()
+        )
+
+        if album:
+            # Add any missing authors
+            current_author_ids = set(album.authors.values_list("id", flat=True))
+            new_author_ids = {author.id for author in authors}
+            missing_authors = new_author_ids - current_author_ids
+            if missing_authors:
+                album.authors.add(*missing_authors)
+            return album
+
+        try:
+            # Create new album
+            album = Album.objects.create(
+                name=album_name, slug=generate_unique_slug(album_name, Album)
+            )
+            album.authors.set(authors)
+            return album
+        except IntegrityError:
+            # Handle race condition
+            album = (
+                Album.objects.select_for_update()
+                .filter(name__iexact=album_name)
+                .first()
+            )
+            if album:
+                album.authors.add(*authors)
+                return album
+            raise
+
+
+def check_song_exists(title: str, album: Album | None, authors: list[Author]) -> bool:
+    """Check if a song already exists with the given title, album and authors."""
+    query = Song.objects.filter(name__iexact=title)
+
+    if album:
+        query = query.filter(album=album)
+
+    if authors:
+        # Ensure exact author match
+        query = query.annotate(author_count=Count("authors")).filter(
+            author_count=len(authors), authors__in=authors
+        )
+
+    return query.exists()
+
+
+def load_mp3_directory(
+    directory_path: str, user_id: int | None = None
+) -> tuple[list[str], int]:
+    """
+    Load all MP3 files from a directory and its subdirectories.
+    Returns tuple of (failed_files, processed_count)
+    """
+    path = Path(directory_path)
+    failed_files = []
+    processed_count = 0
+
+    for mp3_path in path.glob("**/*.mp3"):
+        try:
+            metadata = extract_mp3_metadata(str(mp3_path))
+            if not metadata:
+                failed_files.append(str(mp3_path))
+                continue
+
+            with transaction.atomic():
+                # Process authors
+                author_names = process_authors_string(metadata["artists"])
+                authors = []
+                for author_name in author_names:
+                    author = Author.objects.filter(name__iexact=author_name).first()
+                    if not author:
+                        author = Author.objects.create(
+                            name=author_name,
+                            slug=generate_unique_slug(author_name, Author),
+                        )
+                    authors.append(author)
+
+                # Process album
+                album = None
+                if metadata["album"]:
+                    try:
+                        album = get_or_create_album(metadata["album"], authors)
+                    except IntegrityError as e:
+                        print(f"Error creating album for {mp3_path}: {str(e)}")
+                        failed_files.append(str(mp3_path))
+                        continue
+
+                # Check for existing song
+                if check_song_exists(metadata["title"], album, authors):
+                    print(f"Skipping existing song: {metadata['title']}")
+                    continue
+
+                # Process cover image
+                temp_image_path = None
+                if metadata["image_data"]:
+                    temp_image_path = save_image_as_png(
+                        metadata["image_data"], metadata["image_mime"]
+                    )
+                    if album and not album.image and temp_image_path:
+                        with open(temp_image_path, "rb") as img_file:
+                            album.image.save(
+                                f"{album.slug}.png", File(img_file), save=True
+                            )
+
+                # Create song with proper slug from file name
+                file_name = os.path.splitext(os.path.basename(str(mp3_path)))[0]
+                song = Song(
+                    name=metadata["title"],
+                    length=metadata["length"],
+                    album=album,
+                    slug=generate_unique_slug(file_name, Song),
+                    creator=User.objects.get(id=user_id) if user_id else None,
+                    meta={
+                        "genre": metadata["genre"],
+                        "release_year": metadata["release_year"],
+                    },
+                )
+
+                # Save files
+                with open(mp3_path, "rb") as mp3_file:
+                    song.file.save(f"{song.slug}.mp3", File(mp3_file), save=True)
+
+                if temp_image_path:
+                    with open(temp_image_path, "rb") as img_file:
+                        song.image.save(f"{song.slug}.png", File(img_file), save=True)
+                    os.remove(temp_image_path)
+
+                # Set authors
+                song.authors.set(authors)
+                processed_count += 1
+
+        except Exception as e:
+            print(f"Error processing {mp3_path}: {str(e)}")
+            failed_files.append(str(mp3_path))
+
+    return failed_files, processed_count