Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

LocalLens

github.com

  • Free plan
  • Open source
  • Windows
  • macOS
  • Linux
  • Web
  • Self-hosted
8.3 Club score Visit github.com

What LocalLens is

LocalLens is an offline visual search engine for image and video collections stored on a user's computer. It indexes directories recursively into ChromaDB, then lets users search indexed media with text descriptions and ranks results by similarity. An image upload can find visually similar files, and image and text inputs can be combined to refine a search. A duplicate finder surfaces duplicate and near-duplicate image pairs using a similarity threshold. Listed formats include JPG, PNG, WebP, MP4, and MKV. LocalLens provides a Gradio web interface and a command-line interface, and can sync indexed folders for additions or deletions. Its supported vision embedding backbones include SigLIP 2, MetaCLIP, DFN-CLIP, and CLIP. The project is open-source under Apache-2.0, with a free plan and no paid plans or usage limits stated. Portable builds are available for Windows, Linux, and macOS; Linux and macOS require Python 3.10 or later and Git. Its documented baseline is a 4-core CPU, 8 GB RAM, no VRAM, and 20 MB storage; higher recommended hardware is also listed.

Who it suits

LocalLens suits people who want to search or deduplicate locally stored image and video collections while keeping the tool offline. It offers both a web interface and command-line access for users comfortable with either approach.

Good to know, and what to watch

  • Works offline with locally stored media.
  • Searches by text or uploaded image.
  • Finds duplicate and near-duplicate image pairs.
  • Provides both web and command-line interfaces.
  • Open-source under Apache-2.0.
  • Linux and macOS require Python 3.10 or later and Git.
  • Gallery access defaults to any indexed directory.
  • Recommended hardware includes 4+ GB of VRAM.

Free

Free

LocalLens

Free

Free plan

Getting started

  1. Open the LocalLens project website
  2. Choose a portable build for Windows, Linux or macOS
  3. On Linux or macOS, install Python 3.10 or later and Git
  4. Start the Gradio web interface or command-line interface
  5. Index the directories containing your media
  6. Set LOCALLENS_ALLOWED_PATHS if you want to restrict gallery access

The verdict

LocalLens offers text search, image similarity search, and duplicate finding for local media without an online service. Check its directory access setting and platform requirements before setup.

Questions people ask

How much does LocalLens cost?

LocalLens is free software. The listed LocalLens plan is free.

Which media formats are supported?

The project lists image and video support including JPG, PNG, WebP, MP4 and MKV.

Can I run LocalLens offline?

Yes. LocalLens is described as an offline visual search engine for local collections.

Which platforms are supported?

Supported platforms are Linux, macOS, self-hosted, web and Windows. Portable builds are provided for Windows, Linux and macOS.

Is LocalLens open source?

Yes. It is open-source software licensed under Apache-2.0.

What the maker publishes

Access restriction
The gallery can be restricted to approved directories through the LOCALLENS_ALLOWED_PATHS environment variable. source
Chroma API
Chroma's documentation describes a small Python API for creating collections and adding, updating and deleting records, consistent with LocalLens's ChromaDB indexing layer. source
Community size
The GitHub repository page showed 23 stars, 5 forks and 152 commits when opened. source
Database
LocalLens stores image embeddings in ChromaDB. source
Directory indexing
It recursively indexes directories and supports incremental synchronization for additions and deletions. source
Documentation depth
Official documentation consists of README instructions plus separate installation and hardware-requirements files. source
Duplicate detection
A duplicate finder surfaces duplicate and near-duplicate image pairs using a similarity threshold. source
Duplicate detection
The product identifies duplicate and near-duplicate image and video pairs using a similarity threshold. source
EmbeddingGemma integration
Google's EmbeddingGemma 2 model card describes a 740-million-parameter multimodal embedding model supporting text, images, video and audio. source
Embedding models
Available vision backbones include EmbeddingGemma 2 and several SigLIP 2 variants. source
Gradio role
Gradio describes itself as a Python framework for building machine-learning web interfaces, which is the web UI technology named by LocalLens. source
Hardware baseline
The documented baseline is a four-core CPU, 8 GB RAM, no required VRAM and 20 MB storage. source
Hardware requirements
The documented baseline is a 4-core CPU, 8 GB RAM, 0 GB VRAM, and 20 MB storage; 6+ CPU cores, 12+ GB RAM, and 4+ GB VRAM are recommended. source
Higher-end model cost
The SigLIP 2 512 model is documented as adding about 2.3 GB VRAM, 4 GB RAM and 4.58 GB storage. source
Indexing and sync
It recursively indexes directories into ChromaDB and can sync indexed folders for additions or deletions. source
Install platforms
Portable builds are provided for Windows, Linux, and macOS; Linux and macOS require Python 3.10+ and Git. source
Interfaces
LocalLens provides a Gradio web UI and a command-line interface. source
Interfaces
LocalLens provides a Gradio web interface and a command-line interface. source
Licence
The repository declares the Apache-2.0 licence. source
License
LocalLens is open-source and licensed under Apache-2.0. source
Linux and macOS installation
Linux and macOS manual installs require Python 3.10 or newer and Git. source
Maintainer model activity
The maintainer's Hugging Face profile showed six public models and four followers when opened. source
Maker profile
The maintainer's GitHub profile identifies the account as grinnch and lists LocalLens as a public Python repository. source
Media formats
The README lists image and video support, including JPG, PNG, WebP, MP4, and MKV. source

Checked against the maker's pages on 15 Oct 2026.