GemCollector is an intelligent audio mining workstation. It takes long audio recordings (field recordings, modular jams, vinyl rips) and automatically extracts the musically interesting "gems"—loops, one-shots, and single-cycle waveforms—ready for use in your DAW.
![Screenshots/Demo could go here]
Unlike dumb slicers that chop on fixed grids, GemCollector uses a 3-Pass Heuristic Analysis:
- Interest Map: Scans the file for high-energy and rhythmic sections (ignoring silence).
- Texture Analysis: Calculates sonic characteristics like "Brightness" (Spectral Centroid) and "Percussiveness" (Spectral Flatness).
- Smart Naming: files are automatically named by their content, e.g.,
Loop_120bpm_Bright_Percussive.wav.
- True Stereo: Preserves the full spatial width of your recordings.
- De-Clicking: Applies micro-fades (1ms in / 10ms out) to every slice to ensure clean, click-free audio.
- Non-Destructive: Processing is performed on memory copies, keeping audio pristine.
- Loops: 4-bar segments aligned to the detected BPM.
- One-Shots: Transients (drums, hits) prioritized by musical interest.
- Waveforms: Tiny single-cycle slices for synthesis.
- Download All: Get a timestamped
.ziparchive of your entire session.
The easiest way to run GemCollector is with Docker.
- Docker Desktop installed and running.
Open your terminal in the project root and run:
docker compose up --buildWait a moment for the build to complete.
- Frontend: http://localhost:5173
- Backend: http://localhost:8080
Press Ctrl+C in the terminal, then run:
docker compose downA deployment script is included for Google Cloud Run.
- Google Cloud CLI (gcloud) installed and authenticated.
- A GCP Project created.
Run the script in PowerShell:
./deploy_gcp.ps1 -ProjectId "your-project-id" -Region "us-central1"This will:
- Enable necessary Google Cloud APIs.
- Build Frontend and Backend images.
- Deploy them to Cloud Run (fully managed serverless).
- Output your public production URLs!
- Frontend: React, Vite, Lucide Icons.
- Backend: Python FastAPI, Uvicorn.
- Audio Processing: Librosa, SoundFile, NumPy.
- Infrastructure: Docker, Docker Compose.