VideoSeek
Local video semantic search: locate clips with text or screenshots.
Download VideoSeek
Windows installer (.exe)
Clicking "Download" takes you to a source confirmation page first; the download starts after confirmation.
About VideoSeek
A local-first desktop application for semantic video retrieval, enabling multimodal querying via natural-language text or reference images.
See incorrect information? Tell us
Risk notice
Currently, only a Windows installer is provided. On first run, you must download models, FFmpeg, and other resources yourself; this process depends on the network and takes time. The installer is released by an individual developer, and its digital signature status is not specified in the official information. Before running, make sure you download from the official website (lv17.top) or the official GitHub repository to avoid third-party repackaging. The project is licensed under AGPL-3.0; if you use it commercially or redistribute it, please comply with the corresponding open-source obligations.
Who it's for
- Content creators with large video collections on hard drives or NAS who need to quickly locate clips by visual semantics.
- Editors who need to rough-locate footage before fine editing in an NLE.
- Privacy-conscious users who want video search to run entirely locally without uploading video files.
- Developers who need to expose video search capabilities to agents via a local HTTP API.
Who it's not for
- Users who need native macOS or Linux support (source code requires adjusting the ONNX runtime).
- Users who expect the installation to work without downloading models or FFmpeg separately.
- Users who only need filename-based video management and don't need visual semantic search.
Install guide
Windows Installation Steps
- Go to the official website https://www.lv17.top/ or GitHub Releases to download
VideoSeekSetup.exe(currently only a Windows installer is available). - Double-click the installer and follow the prompts to complete the installation.
- On first launch, the app will prompt you to download and import runtime resources (models, FFmpeg, etc.). Follow the in-app instructions, or manually download and import them via "Settings". You can also download the resources from GitHub Releases (models) or the 123 Cloud Drive zip and import them.
- After importing the resources, add a local folder in "Video Library" and sync videos. Wait for frame extraction and vectorization to finish, then you can search by text or image.
Common Issues
- If Windows SmartScreen shows an "unknown publisher" warning, first verify that the download came from the official website or the official GitHub repository. If it's safe, you can continue via "More info → Run anyway".
- On first use, resource import is required; otherwise, indexing and search will not work.
- The installer and source code share the same UI. If you are using the source code, you can also follow the source-running instructions in the README.
FAQ
Is VideoSeek free software?
Yes, the project is permanently free. The official statement says "permanently free." Do not trust paid resellers or paid installation services. Always download from the official website (lv17.top) or GitHub.
Which operating systems are supported?
The official installer is currently available only for Windows, with Windows 10/11 recommended. On Linux/macOS, you need to run from source and adjust the ONNX runtime yourself (see docs/quickstart.md).
What preparation is needed for first use?
After first launch, follow the in-app prompts to download and import runtime resources such as models and FFmpeg. Then add a folder in "Video Library" and sync videos. Once frame extraction and vectorization are complete, you can search.
Will my video files be uploaded to a server?
No. Video data is indexed and searched locally (ONNX + Lance + FFmpeg), and your video files are never uploaded.
How do I update to the latest version?
Please regularly visit the official website or the GitHub Releases page. When a new version is available, download the installer for that version and overwrite-install. Source users can pull the latest code and build it themselves.