Scan local photos and videos, then find exactly what you need using text descriptions or images. Runs 100% locally β your data never leaves your device.
Powered by the CLIP multimodal model β understands image semantics, not just filenames
Describe what you're looking for in plain language and instantly find semantically matching local photos β no tags required.
Upload any photo and quickly find visually similar images from your local library.
Enter a text description and the system automatically locates and extracts matching video clips, pinpointed to the second.
Upload a screenshot to find the exact video segment it came from β perfect for tracing the source of any clip.
Everything runs locally on your machine. No cloud uploads, no internet connection required, no privacy risk.
GPU acceleration supported. Even on a J3455 CPU, it achieves ~31,000 image matches per second.
Simple setup, instant intelligent search
Use the Windows all-in-one bundle (just extract and run) or deploy with Docker in minutes.
Configure your local photo and video directories. The system automatically scans and builds a smart index.
Open your browser, type a description or upload an image, and find the media you're looking for.
Watch the step-by-step tutorial to get started quickly
Choose the method that fits your workflow
Includes the base model out of the box. Automatically selects GPU acceleration. Perfect for most users.
β GitHub Release DownloadSupports amd64, includes the base model, and supports GPU acceleration. Ideal for Linux servers or NAS devices.
β DockerHub ImageThe core logic is packaged as a pip library, making it easy for developers to integrate or extend.
β materialsearch-coreGot questions? Join the community QQ group or open an issue on GitHub
Join our user community for mutual help and discussion. Group number: 1029566498
β Join QQ GroupFor bug reports, feature requests, or documentation issues, please open a detailed issue on GitHub.
β Open an IssueFree, open-source, and actively maintained β join thousands of users
Go to GitHub β