What is Omoide?

Omoide manages photos and videos on your own hardware without any cloud dependency. Local AI models handle face detection, recognition, and grouping, while OpenCLIP embeddings enable natural-language search across the library. It also finds duplicates, extracts video scenes, plots images on a map, and builds relationship graphs between recognized people. You can run it as a desktop program on Windows, Linux, or macOS, or deploy it in a container on a server or NAS. The code is available under a non-commercial license for personal use.

Should you self-host Omoide?

Omoide scores 40/100 on selfhost.directory (how scoring works) and ranks #37 of 41 in Photo Management. It has 277 GitHub stars.

It shipped 16 stable releases in the past year. The latest, v0.7.6, came out on 21 October 2026. New versions land roughly every 5 days.

Setup is moderate: expect some configuration beyond a single command. An official Docker image is available. It is Other-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is Ente (81/100, 29 702 stars). It scores higher overall, so compare the two before you pick. See Omoide vs Ente side by side.

Features

Auto-Tagging
Background tasks
Co-appearance Graph
Duplicate Detection
Face Recognition
Map View
Multiple Profiles
Read-Only Mode
Semantic Search
Video support

How to configure Omoide

OMOIDE_AI__CLIP_MODELxlm-roberta-base-ViT-B-32OpenCLIP model (options: xlm-roberta-large-ViT-H-14, xlm-roberta-base-ViT-B-32, ViT-L-14, ViT-B-32, convnext_base_w).
OpenCLIP model (options: xlm-roberta-large-ViT-H-14, xlm-roberta-base-ViT-B-32, ViT-L-14, ViT-B-32, convnext_base_w).
OMOIDE_AI__MIN_SEARCH_DIST0.68Minimum cosine distance for semantic search (higher = stricter).
Minimum cosine distance for semantic search (higher = stricter).
OMOIDE_AI__MIN_SIMILARITY_DIST1.2Similarity distance for similar media suggestions. Higher = stronger similarity
Similarity distance for similar media suggestions. Higher = stronger similarity
OMOIDE_DUPLICATES__DUPLICATE_AUTO_HANDLINGkeepHow to handle duplicates (options: keep, remove, blacklist, delete).
How to handle duplicates (options: keep, remove, blacklist, delete).
OMOIDE_DUPLICATES__DUPLICATE_AUTO_KEEP_RULEhighest_resWhich file to keep when auto-handling (options: biggest, smallest, highest_res, lowest_res, oldest, newest).
Which file to keep when auto-handling (options: biggest, smallest, highest_res, lowest_res, oldest, newest).
OMOIDE_FACE_RECOGNITION__CLUSTER_BATCH_SIZE15000Faces processed per clustering batch.
Faces processed per clustering batch.
OMOIDE_FACE_RECOGNITION__EXISTING_PERSON_COSINE_THRESHOLD0.8Cosine threshold for matching faces to existing persons.
Cosine threshold for matching faces to existing persons.
OMOIDE_FACE_RECOGNITION__EXISTING_PERSON_MIN_COSINE_MARGIN0.05Required margin between best and second-best match.
Required margin between best and second-best match.
OMOIDE_FACE_RECOGNITION__FACE_RECOGNITION_MIN_CONFIDENCE0.5Minimum confidence to extract a face.
Minimum confidence to extract a face.
OMOIDE_FACE_RECOGNITION__FACE_RECOGNITION_MIN_FACE_PIXELS1600Minimum face pixel area to be considered.
Minimum face pixel area to be considered.
OMOIDE_FACE_RECOGNITION__HDBSCAN_CLUSTER_SELECTION_EPSILON0.1HDBSCAN cluster selection epsilon.
HDBSCAN cluster selection epsilon.
OMOIDE_FACE_RECOGNITION__HDBSCAN_CLUSTER_SELECTION_METHODeomHDBSCAN cluster selection method (options: leaf, eom).
HDBSCAN cluster selection method (options: leaf, eom).
OMOIDE_FACE_RECOGNITION__HDBSCAN_MIN_CLUSTER_SIZE6Minimum cluster size for HDBSCAN.
Minimum cluster size for HDBSCAN.
OMOIDE_FACE_RECOGNITION__HDBSCAN_MIN_SAMPLES10Minimum samples for HDBSCAN.
Minimum samples for HDBSCAN.
OMOIDE_FACE_RECOGNITION__PERSON_CLUSTER_MAX_L2_RADIUS1Maximum L2 radius allowed inside a face cluster.
Maximum L2 radius allowed inside a face cluster.
OMOIDE_FACE_RECOGNITION__PERSON_MERGE_PERCENT_SIMILARITY75Percent similarity to merge near-duplicate persons.
Percent similarity to merge near-duplicate persons.
OMOIDE_FACE_RECOGNITION__PERSON_MERGE_SEARCH_K20Search K used when merging persons.
Search K used when merging persons.
OMOIDE_FACE_RECOGNITION__PERSON_MIN_FACE_COUNT2Faces needed to auto-create a person.
Faces needed to auto-create a person.
OMOIDE_FACE_RECOGNITION__PERSON_MIN_MEDIA_COUNT2Distinct media items required per person.
Distinct media items required per person.
OMOIDE_FACE_RECOGNITION__PRESETlooseFace clustering preset (options: strict, normal, loose, custom).
Face clustering preset (options: strict, normal, loose, custom).
OMOIDE_GENERAL__ENABLE_PEOPLEtrueEnable face/person features.
Enable face/person features.
OMOIDE_GENERAL__MEME_MODEfalseEnable meme mode labels.
Enable meme mode labels.
OMOIDE_GENERAL__PERSON_RELATIONSHIP_MAX_NODES100Max nodes returned for people graph.
Max nodes returned for people graph.
OMOIDE_GENERAL__PRESENTATION_MODEfalseServe in read-only mode (no writes).
Serve in read-only mode (no writes).
OMOIDE_GENERAL__THUMB_DIR_FOLDER_SIZE1000Max thumbnails per folder.
Max thumbnails per folder.
OMOIDE_PROCESSORS__EXIF_PROCESSOR_ACTIVEtrueEnable EXIF extraction.
Enable EXIF extraction.
OMOIDE_PROCESSORS__FACE_PROCESSOR_ACTIVEtrueEnable face extraction (requires ENABLE_PEOPLE=true).
Enable face extraction (requires ENABLE_PEOPLE=true).
OMOIDE_PROCESSORS__IMAGE_EMBEDDING_PROCESSOR_ACTIVEtrueEnable image embedding generation.
Enable image embedding generation.
OMOIDE_PROCESSORS__MEDIA_BATCH_SIZE125Media rows processed per heavy pipeline batch.
Media rows processed per heavy pipeline batch.
OMOIDE_SCAN__AUTO_CLEAN_ON_SCANfalseClean DB of missing files during auto scan.
Clean DB of missing files during auto scan.
OMOIDE_SCAN__AUTO_CLEANUP_GRACE_HOURS72Grace period (hours) before auto cleanup deletes missing items.
Grace period (hours) before auto cleanup deletes missing items.
OMOIDE_SCAN__AUTO_CLEANUP_WITHOUT_REVIEWfalseRemove missing records without manual review.
Remove missing records without manual review.
OMOIDE_SCAN__AUTO_CLUSTER_ON_SCANfalseRe-cluster faces automatically on scan.
Re-cluster faces automatically on scan.
OMOIDE_SCAN__AUTO_ROTATEtrueRespect EXIF rotation automatically.
Respect EXIF rotation automatically.
OMOIDE_SCAN__AUTO_SCANfalseEnable automatic background scans.
Enable automatic background scans.
OMOIDE_SCAN__IMAGE_SUFFIXES.jpg,.jpeg,.png,.tiff,.bmpImage suffixes to ingest (comma-separated list).
Image suffixes to ingest (comma-separated list).
OMOIDE_SCAN__SCAN_INTERVAL_MINUTES15Minutes between auto scans.
Minutes between auto scans.
OMOIDE_SCAN__VIDEO_SUFFIXES.mp4,.mov,.wmv,.avi,.flv,.mkv,.webm,.gp3,.gif,.ts,.mpegVideo suffixes to ingest (comma-separated list).
Video suffixes to ingest (comma-separated list).
OMOIDE_TAGGING__AUTO_TAGGINGtrueRun automatic tagging.
Run automatic tagging.
OMOIDE_TAGGING__CUSTOM_TAGS,Custom tags (comma-separated list).
Custom tags (comma-separated list).
OMOIDE_TAGGING__USE_DEFAULT_TAGStrueInclude built-in default tags.
Include built-in default tags.
OMOIDE_VIDEO__AUTO_SCENE_DETECTIONtrueUse automatic scene detection for videos.
Use automatic scene detection for videos.
OMOIDE_VIDEO__MAX_FRAMES_PER_VIDEO30Fallback max frames per video when scene detection fails.
Fallback max frames per video when scene detection fails.