PicQuery

English | 中文

PicQuery is an offline image search app. It indexes local folders or photo albums and lets you search them with text or a similar image. Photos, indexes, and model inference stay on your device.

Features include folder and album indexing, Chinese-to-English query translation, image similarity search, folder-scoped search, incremental index updates, and dark mode.

✨ PicQuery V2 has been completely rewritten in Flutter and adds desktop support for Windows, macOS, and Linux. Its database is not compatible with the previous version, so you must rebuild your image indexes after upgrading.

The Play Store version has not yet been updated.

Install

Download the package for your platform from GitHub Releases.

macOS

Because the app is not signed or notarized with a paid Apple Developer account, macOS may prevent it from opening. To allow it, open Terminal and run:

sudo xattr -rd com.apple.quarantine "/Applications/PicQuery.app"

Build from source

Install Git, Git LFS, Flutter 3.47.4 (stable), and the Flutter toolchain required by your target platform.

git lfs install
git clone https://github.com/greyovo/PicQuery.git
cd PicQuery
git lfs pull
flutter pub get

Before running or building the app, assets/models/ must contain:

• mobileclip2_s0_visual.onnx
• mobileclip2_s0_text.onnx
• mt_zho-eng.fp32.quantized.onnx // Stored with Git LFS

Models are exported by export.sh in greyovo/picquery-models. Install uv, then run:

bash scripts/prepare_models.sh

The script uses the sibling directory ../picquery-models by default. If missing, it clones the repository, installs dependencies, exports models, and copies them into assets/models/. ci/build.sh runs this preparation automatically before building the app.

Run the app:

flutter run

# For lint and tests
flutter analyze lib/
flutter test

Contributing and acknowledgments

Issues and pull requests are welcome. Please include reproduction steps, the affected platform, and relevant check output. Model or preprocessing changes should include numerical and retrieval validation.

PicQuery builds on Apple MobileCLIP. Thanks to @mazzzystar and @Young-Flash for their help; see the original discussion.

• mazzzystar/Queryable, the inspiration for this project and an iOS app
• IacobIonut01/Gallery

License

The project source is available under the MIT License. Third-party dependencies and model assets retain their respective licenses, including the Apple MobileCLIP model license.