For a very good description and detailed instructions see Logix's Article at Linux Uprising
Google Photos Sync downloads your Google Photos to the local file system. It will backup all the photos the user uploaded to Google Photos, but also the album information and additional Google Photos 'Creations' (animations, panoramas, movies, effects and collages).
This project uses the new Google Photos API see https://developers.google.com/photos/.
After doing a full sync you will have 2 directories off of the specified root:
- photos - contains all photos and videos from your Google Photos Library organized into folders with the structure 'photos/YYYY/MM' where 'YYYY/MM' is the date the photo/video was taken. The filenames within a folder will be as per the original upload except that duplicate names will have a suffix ' (n)' where n is the duplicate number of the file (this matches the approach used in the official Google tool for Windows).
- albums - contains a folder hierarchy representing the set of albums and shared albums in your library. All the files are symlinks to content in the photos folder. The folder names will be 'albums/YYYY/MM Original Album Name'.
In addition there will be further folders when using the --compare-folder option. The option is used to make a comparison of the contents of your library with a local folder such as a previous backup. The comparison does not require that the files are arranged in the same folders, it uses meta-data in the files such as create date and exif UID to match pairs of items. The additional folders after a comparison will be:
- comparison a new folder off of the specified root containing the following:
- missing_files - contains symlinks to the files in the comparison folder that were not found in the Google Photos Library. The folder structure is the same as that in the comparison folder. These are the files that you would upload to Google Photos via the Web interface to restore from backup.
- extra_files - contains symlinks into to the files in photos folder which appear in the Library but not in the comparison folder. The folder structure is the same as the photos folder.
- duplicates - contains symlinks to any duplicate files found in the comparison folder. This is a flat structure and the symlink filenames have a numeric prefix to make them unique and group the duplicates together.
NOTES:
- the comparison code uses an external tool 'ffprobe'. It will run without it but will not be able to extract metadata from video files and revert to relying on Google Photos meta data and file modified date (this is a much less reliable way to match video files, but the results should be OK if the backup folder was originally created using gphotos-sync).
- If you have shared albums and have clicked 'add to library' on items from others' libraries then you will have two copies of those items and they will show as duplicates too.
- Installing on a slow machine (like old Raspberry Pi) or network may cause timeouts in pipenv.
- This can be resolved by setting an environment variable export PIPENV_TIMEOUT=240
- Some mounted filesystems including NFS, CIFS and AFP do not support file locks and database access will fail on them.
- To fix, use the paramter --db-path to sepcify a location for your DB on the local disk. This will perform better anyway.
A few outstanding limitations of the Google API restrict what can be achieved. All these issues have been reported to Google and this project will be updated once they are resolved.
- There is no way to discover modified date of library media items. Currently
gphotos-syncwill refresh your local copy with any new photos added since the last scan but will not update any photos that have been modified in Google a. Photos. A feature request has been submitted to Google. - FIXED BY GOOGLE. Some types of video will not download using the new API.
- The API strips GPS data from images.
- Video download transcodes the videos even if you ask for the original file (=vd parameter). My experience is that the result is looks similar to the original but the compression is more clearly visible. It is a smaller file with approximately 60% bitrate (same resolution).
- Burst shots are not supported. You will only see the first file of a burst shot.
For an easy option which does not require the install of Python and Pipenv you can use the Snap Store version, see https://ubuntu.com/blog/safely-backup-google-photos.
For some help on getting python working on Windows see gilesknap#63.
On linux, you can install pipenv using pip3 install --user pipenv and then make sure that ~/.local/bin/ is in your path.
To install the latest published version from PyPi, simply:
mkdir gphotos-sync cd gphotos-sync pipenv install gphotos-sync
Or if you don't want to use pipenv, create a virtual environment and:
pip install gphotos-sync
(see https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/ if you are not familiar with virtualenv)
To work from the source code, clone the git repository and use pipenv to create a virtual environment and run the code. (if you don't have pipenv, then I recommend getting it - but you can manually create a virtualenv and use 'python setup.py install' instead)
git clone https://github.com/gilesknap/gphotos-sync.git cd gphotos-sync pipenv install . pipenv run gphotos-sync
In order to work, gphotos-sync first needs a valid client id linked to a project
authorized to use the 'Photos Library API'. It is not provided in the distribution. Each client id
is given a (large) limited number of free API calls to Google Services. If this distribution shared the client id,
all users would share this resource limit. This is a little fiddly but only needs to be done once.
- Create a client id using these instructions Oauth2 for gphotos-sync
- Once the client ID is created, download it as
client_secret.jsonand save it under the application configuration directory:~/Library/Application Support/gphotos-sync/under Mac OS X,~/.config/gphotos-sync/under Linux,C:\Users\<username>\AppData\Local\gphotos-sync\gphotos-sync\under Windows.
Also note that for Windows you will need to enable symbolic links permission for the account that gphoto-sync will run under. See Enabling SymLinks on Windows.
Once the script is configured, you are now ready to use it using the simple following command line:
cd <installed directory> pipenv run gphotos-sync TARGET_DIRECTORY
Or, if you used virtualenv and pip instead of pipenv, activate the virtualenv and:
gphotos-sync TARGET_DIRECTORY
The first time, it will give you a link to an authorization page in order to authorize the client to access your Google Photos.
For a description of additional command line parameters type:
gphotos-sync --help
You can run the tool from the container using prebuilt Docker image. The container has 2 mount points:
/storagethis is where your photos will be stored. You can mount single directory, or multiple subdirectories in case you want to backup multiple accounts/configthe directory that contains client_secret.json file
To run
docker run \ -ti \ --name gphotos-sync \ -v /YOUR_LOCAL/PATH/TO_PHOTOS:/storage \ -v /YOUR_LOCAL/PATH/TO_CONFIG:/config \ gilesknap/gphotos-sync \ /storage
To remove the container (for instance if you want to run it on scheduled basis and do a cleanup):
docker rm -f $(docker ps --filter name=gphotos-sync -qa) 2> /dev/null
To run then remove the container:
docker run \ --rm \ -it \ --name gphotos-sync \ -v /YOUR_LOCAL/PATH/TO_PHOTOS:/storage \ -v /YOUR_LOCAL/PATH/TO_CONFIG/client_id.json:/config/client_secret.json:ro \ gilesknap/gphotos-sync \ --log-level INFO \ /storage
I have just experienced an issue with duplication of files when doing a rescan (--rescan or --flush-index). It looks like some items have changed in the library and this can result in the same file downloading twice. I would guess this has something to do with Google removing the Drive link to Photos.
UPDATE: I now know that this was caused by subtle changes in the metadata. It seems Google does not guarantee to deliver exactly the same files each time you scan the library (but to be fair, I think they are tuning things for the better).
The problem did cause some duplicate named files to be downloaded twice overwriting their duplicate peer. Note that no files were lost from the library (since gphotos is read-only) and it was possible to repair things by either:
- using the local comparison feature of gphotos-sync against a prior backup
- or downloading the library again from scratch
In summary, most people will not be affected by the issue I had unless they have very old photos with duplicate file names.
My detailed notes on the subject are here: giles notes
UPDATE: the GPS scraping no longer works and has been removed. I am investigating a couple of other avenues.
Google does not seem to be interested in fxing the issue of stripping location info from the EXIF info of images downloaded via their API (see https://issuetracker.google.com/issues/80379228#comment80). So I am investigating a workaround. See the option --get-locations. It uses Selenium to scrape the GPS info off of the Google Website (your google creds required I'm afraid) and insert them into the DB of synchronized files. It does not yet update the EXIF on the local files but this is a minor addition and I'll implement if there is interest.
Have a try and let me know what you think. Hurry, because Google is removing the ability to log in using automation soon!