Home
Culvia is a local-first photo culling, scoring, review, and export workbench for photographers.
It helps turn a full shoot into a smaller, intentional set of images by combining folder scanning, thumbnail generation, local aesthetic and technical scoring, optional OpenAI-compatible vision review, manual decisions, filtering, and export tools.
Photo culling is repetitive, visual, and judgment-heavy. Culvia is built to make that first selection pass faster while keeping the photographer in control.
The screenshots below use anonymized photo content and redacted local paths.
The viewer keeps the current photo, thumbnail strip, scoring summary, manual decisions, and accept actions in one place.
The photo wall is for scanning a filtered set, comparing scores, and applying quick culling decisions.
Detailed metrics live below the thumbnail strip, with large-model scores and comments split into readable columns.
The app is useful when you want to:
- Review a large folder of photos without importing into a hosted service.
- Sort by recommendation, technical quality, large-model review, manual status, or color label.
- Compare local scoring with optional image critique from an OpenAI-compatible vision model.
- Keep manual picks, holds, rejects, star ratings, and color labels as the final curation layer.
- Export selected images or a CSV review table for downstream editing and delivery.
Culvia is an early-stage open-source project. The core web workflow is usable, and the desktop shell and release packaging are still evolving.
The project is intentionally local-first: runtime databases, thumbnails, uploads, exports, and local model caches stay on the user's machine by default.
- Repository: https://github.com/yangzhg/culvia
- English README: https://github.com/yangzhg/culvia#readme
- Chinese README: https://github.com/yangzhg/culvia/blob/main/README.zh-CN.md
- Issues: https://github.com/yangzhg/culvia/issues