Culvia
github.com
What it is
Culvia is a local-first workbench for photographers to cull, score, review, and prepare photos for delivery. It can import folders or temporary uploads, scan subfolders, and deduplicate repeated paths. Local models score images for aesthetic and technical qualities and generate thumbnails for browsing. An optional OpenAI-compatible vision review can add image critique, subscores, retouching advice, and shooting advice. Photographers make the final selections by marking images as picks, holds, or rejects, with star ratings and color labels also available. Filters and sorting cover recommendations, technical quality, AI review, manual status, and other review dimensions. Culvia exports picked photos and CSV results for later editing and delivery; documented CSV fields map to Lightroom ratings, flags, and color labels, and to Capture One ratings and color tags. Photos and working data stay on the user's machine by default; images go to an external service only when vision review is enabled. The project is free and open source, with desktop packages for macOS, Windows, and Linux and a local web app. It is early-stage and does not replace a full RAW editor.
Spec sheet
- What it does
- Culvia is a local-first photo culling, scoring, review, and delivery workbench for photographers.Read 9 Oct 2026 · github.com
- Purpose
- Culvia is a local-first photo culling, scoring, review, and delivery workbench for photographers.Read 8 Oct 2026 · github.com
- Intended users
- It is aimed at photographers reviewing similar frames from portraits, events, weddings, street sessions, practice shoots, editorial selects, and personal archives.Read 9 Oct 2026 · github.com
- Platforms
- The v0.1.0 release lists desktop packages for Linux, Windows, and Apple Silicon macOS; the README also describes a local web app.Read 8 Oct 2026 · github.com
- Integrations
- Exported CSVs include helper columns mapped to Lightroom and Capture One ratings, flags, and color labels; writing XMP, sessions, catalogs, or sidecar files is described as future work.Read 9 Oct 2026 · github.com
- Privacy
- Local models, thumbnails, previews, SQLite data, uploads, and exports stay on the user's machine by default, and photos are sent to an external service only when the user explicitly enables OpenAI-compatible vision review.Read 9 Oct 2026 · github.com
- AI review
- An optional OpenAI-compatible vision review can provide image critique, aesthetic and technical sub-scores, retouching advice, and shooting advice.Read 9 Oct 2026 · github.com
- Credential handling
- The README says API keys should be entered through app configuration or environment variables, not placed in documentation, tests, logs, plaintext SQLite fields, or Git.Read 9 Oct 2026 · github.com
- Credentials
- The README says API keys should be entered through app configuration or environment variables, not stored in documentation, tests, logs, plaintext SQLite fields, or Git.Read 8 Oct 2026 · github.com
- Desktop security
- The v0.1.0 release says it includes SHA-256 checksum files and GitHub Artifact Attestations, with a command for verifying release asset provenance.Read 8 Oct 2026 · github.com
- Desktop signing
- The README says current desktop binaries are not Developer ID signed, notarized, or Windows code signed, and may trigger operating-system warnings.Read 9 Oct 2026 · github.com
- Exports
- Culvia can export picked photos and CSV results for downstream editing and delivery workflows.Read 9 Oct 2026 · github.com
- Filtering
- Users can filter and sort by recommendation, technical quality, LLM review, manual status, color label, and other review dimensions.Read 8 Oct 2026 · github.com
- Limits and status
- Culvia describes itself as an early-stage open-source project and says it is not a replacement for a full RAW editor such as Lightroom Classic, Capture One, or darktable.Read 8 Oct 2026 · github.com
- Local scoring
- Culvia can run local aesthetic and technical scoring models and generate thumbnails for gallery browsing.Read 9 Oct 2026 · github.com
- Manual review
- Users can mark photos as pick, hold, or reject, assign star ratings and color labels, and filter or sort by review dimensions.Read 9 Oct 2026 · github.com
- Photo handling
- Users can import photo folders or temporary uploads, scan subfolders, and deduplicate repeated paths.Read 8 Oct 2026 · github.com
- Project maturity
- Culvia is described as an early-stage open-source project whose workflow, desktop packaging, and model integrations are still maturing.Read 9 Oct 2026 · github.com
- Release verification
- The v0.1.0 release includes SHA-256 checksum files and GitHub Artifact Attestations for release assets.Read 9 Oct 2026 · github.com
- Scope limit
- The project says it is not a replacement for a full RAW editor such as Lightroom Classic, Capture One, or darktable.Read 9 Oct 2026 · github.com
- Scoring
- Culvia can run local aesthetic and technical scoring models.Read 8 Oct 2026 · github.com
- Supported systems
- The project publishes desktop packages for macOS, Windows, and Linux and also offers a local web app through its Python package.Read 9 Oct 2026 · github.com
- Workflow
- It combines local scoring, optional vision-model critique, manual pick/hold/reject decisions, and export tools.Read 8 Oct 2026 · github.com
Strong points and limits
Strong points
- Local models score aesthetic and technical qualities.
- Manual selections remain authoritative.
- Exports picked photos and CSV results.
- Photos stay local unless vision review is enabled.
- Desktop packages and a local web app are available.
Limits
- The project is early-stage and still maturing.
- It is not a replacement for a full RAW editor.
- Desktop binaries are not signed or notarized.
Who it suits
Culvia suits photographers reviewing similar frames from portraits, events, weddings, street sessions, practice shoots, editorial selects, or personal archives. It is for culling and review workflows rather than full RAW editing.
Verdict
Culvia combines local scoring, optional external vision review, manual selection, and export in a focused photo-culling workflow. Its early-stage status and unsigned desktop binaries are worth weighing before adoption.
Where Specifiction places it
- Photo Culling Software#18 of 267.7