immich vs ente photos - the photo backup showdown
59 points by raymii
59 points by raymii
It's astounding how good immich is, let alone for something available to us at no cost
bravo to the creators/maintainers putting so much effort into this project
i just found immich recently. very impressive with the quality of the project.
sync is super fast. i used to the slow speed of synology photos and immich speed is refreshing.
With a title like that I was really hoping to get a perspective on the actual day to day usability of ente, or even a look into how it is to operate it over time. I mostly see people talking about immich, and even run it myself, but I'm really curious about ente
it is very usable, sharing with other people works great, restoring works great (just install the app and sign in). one very neat feature is that it builds a map: https://imgur.com/4RWfdEL
(don't worry it becomes more fine grained when you zoom in)
the only annoyance i have that face detection and grouping takes a while
I've found the face detection to be pretty lacking. I like the company a lot so I deal with it but I don't share smart albums because of how frequently it misrecognizes people.
post reads like someone turned off auto-caps on an AI summary. some good info regardless
Pretty likely. pangram (which is the only low false positive detector option) thinks it's entirely ai too.
Could be an AI rewrite from an outline/draft, could be an ex nihilo gen. Hard to tell.
https://www.photoprism.app/ Is a another great option
I use both - Immich for my wife and I’s iphone photo backups so I can ditch iCloud and then I use Photoprism as more of a photography CMS for the pictures I take with my actual digital and film cameras and other general scanned archival media from my family and stuff I want to preserve. For some reason I’ve found their usage subtly different in this way and I’m grateful it works to my benefit, I’m sure I’m somewhat unique in this way.
Immich Is fine on a lower resource server if you're not running the ML stuff on the same machine; you can run the ML service on a separate machine and point Immich to it.
I run it on my overloaded raspberry pi 4 along with a dozen other things, and the ML service on my desktop. The only annoyance is that search by keywords doesn't work when the desktop is not running and that I have to manually log in to run the queued jobs every now and then.
The ML stuff is bursty too. So at idle I'm using about 1GiB of RAM and basically no CPU. Not tiny but if you have a decent box that runs a few other things the burst workload will probably be in the background. On NixOS at least it also runs in a different systemd service so you can easily change the priority of the machine learning stuff so that it doesn't get in the way of other workloads.
If I had to estimate you could probably get by on 2GiB RAM total if you are willing to turn down the ML stuff a bit (smaller models, lower OCR resolution) and would still have a pretty good experience.
You can still run the ML on the same machine. I run Immich fully on a 4core Atom and it takes a while to process photos that are uploaded, but the keyword/similarity search is plenty fast.
I've been looking to set up Ente, and eventually I'll do that, no need to remind me every 6 months :) I'm wondering would it be possible to disable at-rest encryption?
Can you turn off the face-scanning features? If so, does that vastly decrease system requirements? I just want something simple.
Yeah, you can choose which machine learning components to disable: https://docs.immich.app/administration/system-settings/#machine-learning-settings
Edit: The docs aren't entirely clear on it, but to disable them you go to the administration interface inside Immich and disable either the entire machine learning feature, or individual components (like facial recognition).
According to their docs, disabling all machine learning features lowers the recommended amount of RAM from 8 to 4 GB.
Alternatively one can offload the machine learning to another machine, like your probably much more powerful laptop or desktop: https://docs.immich.app/guides/remote-machine-learning
So I'm very interested in these, but my question is, how well do the iOS apps work? Or, is there a shim that runs on a Mac? I really value the "fire and forget" nature of Apple's photo syncing, even as I'd like to run it myself.
I’ve only tried immich, but the iOS app works very well. It periodically wakes up and syncs and pending photos. Has been working seamlessly for my wife and I for the past couple of months.