Despite skepticism I love to see experiments like that. If we all are able to run an open source model locally on mid-high end machines I'd be very happy.
I use my Immich instance locally as photos and videos storage for 7-8 months now. It's really amazing. Reading comments here I'd say it's medium-difficulty tier to set up. I have it on my Elitedesk 800 G3 mini, in Docker Compose, with hardware encoding and it's more than sufficient to handle this work. Never had a bug.
I fail to get the problems with self hosting Immich. I mean, obviously you gotta have a server ready, and at least some knowledge about self hosting and docker. But apart from that, installing Immich was like a breeze. It was actually not more than pasting some lines into a docker-compose, and running it.
Have been using it for about 1.5 years, and I have not a single problem, which is quite incredible for a software that basically has all features that Google Photos has.
As a counter-anecdote: literally just `docker compose up` on a spare laptop for me, and it's working great (though it's only available on the local network). There might be stuff to tune (e.g. I'm pretty sure it's not using my GPU), but it's almost totally unnecessary for just one household of people's use - the initial huge google-photos-takeout took an hour or three to finish indexing with all the features enabled, but all new stuff is done within seconds. The most I've done is to swap the actual photo storage to an external drive, which is just a "move that folder, custom mapping in the docker command" change.
On hardware that doesn't have docker, or is significantly more resource constrained somehow: yea, I completely believe it. I haven't tried that, but given the features it makes total sense that it'd be harder.
I posted this blog entry mostly for my own reference (I use my blog as a searchable index of all my findings, since Google is easier to use than any note taking app), and it always surprises me what makes it into HN vs what doesn't...
I think in this case it may have unintentionally riled up the "Ubuntu isn't what it used to be" sensibility that's prevalent among Linux users, but I see it more as an annoying little bug that will hopefully be fixed in stable releases soon... but until then, here's the workaround.
It's pretty par for the course for the Gnome team. Gnome 3 had a major impact usability on every single distro, and there were many more problems following that.
The Ubuntu team probably doesn't feel safe updating the Settings app without extensive regression testing.
So we're building binaries, to run them in WASM, that's running in JavaScript, which is running in browser, that is written in C++ .. is that what we're doing now?
reply