Curlo
Local AI search to find SFX and music by describing it
54 followers
Local AI search to find SFX and music by describing it
54 followers
Curlo is a privacy-first macOS app for searching, previewing, and organizing large sound libraries. Find SFX or music by describing what you want to hear, search for similar sounds, edit metadata & UCS, manage tags, and keep everything fully local on your Mac.
Another Dock
Hey Product Hunt 👋
I’m the maker of Curlo, and also a part-time audiobook post-production editor.
Curlo started from a problem I kept running into: sometimes you know what you want to hear, but you do not know the filename.
So I built Curlo — a native macOS app that lets you search your local sound library by describing what you hear.
For example:
- heavy rain on a tin roof
- footsteps on gravel
- gentle piano melody
- short metallic hit
- dark cinematic drone
Curlo analyzes the audio content locally on your Mac, then lets you find sounds by natural language. The search does not need the file to already be named or tagged perfectly. Your audio files and search queries stay on-device.
Key features:
- Semantic search for local audio files (Fully offline)
- Metadata search with commands like /filename, /tag, /ucs, and /format
- Drag clips into DAWs and apps like Pro Tools, Logic Pro, Reaper, Premiere Pro, and other tools
- Similar audio search from a file or selected region
- AI auto-tagging and UCS category support
- Metadata reading, editing, and writing
- Localhost API for custom workflows
I’d love feedback from sound designers, video editors, audiobook/podcast editors, music producers, and anyone managing a large local audio library.
Thanks for checking it out!
Describing a sound and getting matches while staying fully local implies an on-device audio embedding model doing the heavy lifting. How big can a library get before search starts to drag — does it index everything once on import, or scan on each query?
Another Dock
@sounak_bhattacharya Curlo builds the index once during import. If files are modified later, users can manually trigger a rescan, and Curlo will automatically re-index only the files that have changed rather than rebuilding everything.
Based on user feedback so far, even libraries around 100,000 audio files have not had a noticeable impact on search speed.
Very cool use case for AI. Curious how well it handles niche sound requests?
Another Dock
@nithin_raju1 Good question. I'm still exploring the edge cases. In general, the more specific the prompt is about the scene, material, action, and mood, the better AI performs. If you have a niche sound in mind, I’d love to test it.