Indexing Folders

Indexing is the process by which SoundBakery scans your audio folders, tags each file with AI, and builds a searchable database. This page explains how it works and how to manage your indexed libraries.

Adding Folders

To index a folder of audio files:

1. Open the Sidebar

   Click the folder icon in the left sidebar. If collapsed, hover over the left edge of the window to reveal it.

2. Click Add Folder

   Click the Add Folder button at the top of the folders sidebar. A standard macOS file picker opens.

3. Select a Directory

   Navigate to and select any folder containing audio files. The app will scan recursively (including subfolders).

4. Wait for Indexing

   Indexing begins automatically. Progress is shown in the sidebar and a status indicator. You can continue using the app while indexing runs.

The Indexing Pipeline

SoundBakery processes each folder through a multi-stage pipeline:

Pass	Stage	What Happens
1	Catalog Scan	The app recursively scans the folder listing all audio files. For each file it reads: filename, file size, last modified date, duration, sample rate, channel count, and format. This data is stored in a SQLite database.
2	CNN Audio Tagging	Each audio file is processed through a CNN classifier (either the bundled Honeycake model or Apple's SoundAnalysis). The model predicts acoustic characteristics and generates tags like "bright," "reverberant," "impact," "texture," etc.
3	LLM Tag Expansion	The raw CNN tags are expanded by the LLM into natural-language descriptions. For example, tags like "impact_metal_high" become "A sharp metallic impact sound with high-frequency resonance — suitable for hits, clanks, and industrial effects."
4	Embedding Generation	The expanded descriptions are converted into semantic vectors (embeddings) by the embedding model. These vectors enable natural-language search — when you search for "warm sub bass," the embedder converts your query into a vector and finds the closest matches.

Indexing Behavior

• New files — Automatically detected and indexed on re-index.
• Modified files — Detected by changed modification timestamps. Re-indexed on next scan.
• Deleted files — Removed from the catalog on next full re-index.
• Pause/Resume — Indexing can be paused and resumed via the sidebar controls.
• Background indexing — Continues even if you switch to other tasks in the app.

Supported File Types

SoundBakery supports common audio file formats:

Format	Extension	Notes
WAV	.wav	Uncompressed PCM — fully supported
MP3	.mp3	Compressed — supported for playback and indexing
AIFF	.aiff, .aif	Uncompressed — fully supported
CAF	.caf	Apple Core Audio — fully supported
FLAC	.flac	Lossless compressed — supported for playback
M4A/AAC	.m4a, .aac	Compressed — supported

Managing Indexed Folders

Removing a Folder

Right-click a folder in the sidebar and select Remove Folder. This removes the folder from the index but does not delete the original files.

Re-indexing

Click the refresh icon next to any folder to trigger a re-scan and re-index. This picks up new, modified, and deleted files.

Pause and Resume

During indexing, a pause button appears in the sidebar. Click it to pause indexing; click again to resume where it left off.

Database Maintenance

Over time, the indexing database may accumulate stale data. SoundBakery includes maintenance tools:

• Clean Index — Removes entries for files that no longer exist on disk. Also removes orphaned embedding vectors.
• Rebuild Vector Index — Rebuilds the sqlite-vec approximate nearest neighbor index for faster search.
• Backup Database — Creates a backup copy of the catalog database (catalog.sqlite).
• Restore Database — Restores from a previous backup.

Access these tools from the File → Database Maintenance menu.