An AI-powered desktop sample manager built for music producers. Visit website →
SampleFind turns large, disorganized sample folders into a searchable production library. Producers can scan their sample folders, search by using natural language, instrument, category, sound, key, BPM, or sample type, preview results instantly, and drag audio directly into their DAW.
Downloads are provided for portfolio review and controlled testing. They are not currently notarized, code-signed production releases.
Watch the complete SampleFind demo →
Music producers often accumulate thousands of samples across folders and sample packs. File names are inconsistent, categories overlap, and manually searching through multiple large folder for the right sound can be inconvenient and can interrupt the creative process.
SampleFind creates a local-first index and combines filename heuristics, audio
metadata, and semantic audio analysis to deliver focused results for both broad
queries such as Middle East and specific searches such as
darbuka, kick one shot, high hat loop, or A minor melody.
- Local-first indexing — original audio files remain in the folders selected by the user.
- Hybrid search — combines filename signals, categories, metadata, and audio similarity.
- Producer-focused filters — category, one shot, loop, fill, key, BPM, and favorites.
- Background scanning — persistent progress, queued folders, cancellation, rollback, and completion reporting.
- Audio workflow — instant previews, global playback controls, and drag-and-drop into supported DAWs.
- Source management — rescan or remove indexed folders without deleting the original files.
- Cross-platform packaging — versioned macOS and Windows desktop installers.
- Account experience — authentication, profile management, and password recovery.
flowchart LR
A["User-selected audio folders"] --> B["Background scanner"]
B --> C["Metadata + audio analysis"]
C --> D["Local indexed library"]
D --> E["Hybrid search and layered filters"]
E --> F["Preview, favorite, or drag into DAW"]
- Audio–text AI: LAION-CLAP embeddings connect natural-language queries with the sound of an audio sample, even when filenames are incomplete.
- Hybrid retrieval: semantic similarity is combined with filename heuristics, structured metadata, category signals, and ChromaDB vector search.
- Intelligent classification: model-assisted audio understanding works alongside high-confidence rules for instruments, drum parts, loops, one shots, and fills.
- Audio intelligence: Librosa-based analysis enriches samples with BPM and musical-key metadata, with filename metadata taking priority when available.
- Local data layer: SQLite manages application metadata, source history, classifications, favorites, and user-specific corrections, while ChromaDB powers local embedding retrieval.
- Product stack: Next.js, React, TypeScript, Tailwind CSS, FastAPI, Python, Electron, and Firebase authentication.
- Release engineering: GitHub Actions, macOS DMG packaging, and Windows NSIS distribution.
SampleFind demonstrates end-to-end product development across:
- Applied multimodal AI — translating producer language into audio–text embeddings that can retrieve samples by what they sound like, not only what they are named.
- Hybrid relevance ranking — prioritizing exact filename and metadata matches while using semantic similarity for broader or unlabeled searches.
- Query-aware precision — treating specific instrument searches more strictly while allowing wider instrument families and regional categories to return appropriately broader results.
- AI-assisted taxonomy design — organizing instruments, drum parts, effects, regional sounds, and sample types into searchable classifications.
- Human-in-the-loop correction — allowing user-specific manual reclassification without silently retraining or changing global model behavior.
- Evaluation-driven development — maintaining representative query sets for exact instruments, sound families, genres, BPM, key, aliases, and ambiguous producer terminology.
- Latency-conscious local inference — keeping search and metadata retrieval local while balancing semantic analysis with fast deterministic signals.
- Reliable background processing — scan queues, persistent progress, cancellation, transactional rollback, and publish-on-completion behavior.
- Full-stack product engineering — responsive interfaces, authentication, local filesystem integration, cross-platform packaging, and automated releases.
This public repository is the home of the SampleFind product showcase, public documentation, and future landing page. The production desktop application source is maintained in a private repository. Source access can be provided to technical interviewers when appropriate.
Important
Portfolio beta — not a publicly supported consumer release. SampleFind is a functional cross-platform beta shared for portfolio review and controlled evaluation. The current macOS build is not Apple-notarized and the Windows installer is unsigned, so operating systems may display security prompts. The product is architected and packaged for public deployment, but a public launch would additionally require production code signing and notarization, final distribution QA, privacy/legal documentation, and a supported update channel.
Built by Krishy.