No description
Find a file
2026-02-12 21:46:30 +01:00
frontend Frontend code cleanup, remove dead code 2026-02-11 22:17:26 +01:00
resources Update readme and setup docker image tagging 2025-04-13 14:26:50 +02:00
src Revert some changes that made the pipeline slower. 2026-02-12 21:46:30 +01:00
.gitignore Frontend code cleanup, remove dead code 2026-02-11 22:17:26 +01:00
Cargo.toml small improvements to the rust backend 2026-02-12 20:20:12 +01:00
LICENSE Add license 2025-03-29 09:36:12 +01:00
README.md Update readme and setup docker image tagging 2025-04-13 14:26:50 +02:00

Immich Selfie Timelapse Tool

This tool helps create selfie timelapses from your Immich instance.
It uses the powerful machine learning features of Immich to gather all the photographs where a particular individual appears, retrieves the bounding box metadata, and automatically crops and aligns the photos.

Example Image

Example GIF

Features

  • Automatically fetch images featuring a specified individual from your Immich instance.
  • Extract bounding box metadata and crop/align photos using machine learning.
  • Discard photos with low resolution (set by threshold).
  • Discard photos where the subject is viewed from the side.
  • Adds timestamp in the filename for easy chronological ordering.

Docker compose setup

services:
  immich-selfie-timelapse:
    image: arnaudcayrol/immich-selfie-timelapse
    container_name: immich-selfie-timelapse
    ports:
      - "5000:5000"
    volumes:
      - ./immich_selfie_timelapse:/app/output
    environment:
      - IMMICH_API_KEY=abcdefghijklmnopqrstuvwxyz
      - IMMICH_BASE_URL=http://192.168.1.94:2283/api

Once the service is started, access the webpage to configure the tool: http://127.0.0.1:5000.

Example GIF

Additional Notes

  • For best looking results, I still recommand a manual pass over the final image folder to remove some outliers. The face orientation detection is not perfect at all. Hopefully someday I can get a better results.
  • Contributions and improvements are welcome.

License

This project is open source and available under the MIT License.