And the thing I particularly enjoy about immich is that it's semantic searching is leagues better than any of the proprietary solutions in a lot of use cases. Particularly it works amazing for various brands and IP that 'cloud' offerings simply wont recognize.
Most people would not need the storage they are using if the just took less photographs and edited more, but good on Immich.
Followed by 300 lines of obscure config commands and server configuration. Great...
My solution: I connect my iPhone once or twice a year and import my photos to a folder on my hard drive. Then I delete the old photos from the phone. That has worked fine for a couple of decades now.
Or you can use iCloud or Google Photos, which is about one dollar per month.
Your second solution is much closer to Immich in terms of feature set, but (AFAIK) costs a bit more than 1$/month, depending on how much data you have. And it also doesn't match the privacy you'd get with Immich. Google will happily run your private photos through whatever AI model they feel like, without your permission or consent.
I'm not saying your solutions aren't good, if it works for you then great! But they're not really something you can compare to having an Immich instance
Your solution, as presented, assumes you don’t lose any photos in six months and never go above the 2gb or so of space you get for a dollar. The impetus to finally set up Immich came from my wife pushing past that. Even at fifty cents a GB forever, it’s not going to scale as well as essentially 0, though that depends on considering the drive as a sunk cost.
And now I finally have to suck it up and establish an off-site backup plan to make my scenario like for like. Damn you!
I built malmo to solve that. 1 click install for immich malmo.network/store/immich
Underlying OS fully open source, install it on your own box or use a hosting provider
What are you trying to communicate to us? Maybe you are just trying to tell people that they can get away with simpler solution if all they care about is unscheduled photo backup. If so, I feel there is a less antagonistic way for you to say it.
Otherwise I think Immich seems like a cool project, and I hope they're successful.
I added the part about putting photos in a folder, to at least contribute with a different solution, which for me has worked very well for very long.
I do both, folders for backup, Immich for indexing and mobile access. Real simple docker compose file I wrote once to get the whole thing going years ago.
Doesn’t have to be either or, and a dumb folder only accomplishes a portion of my needs.
Unless you actually go through the "Backing up your photos" step, you _will_ lose your photos, it's just a matter of time. And if you're like most people, that is probably not acceptable.
And by the time you do the math for the cost of your backups, you'll find Google Photos pricing is actually pretty good.
Still worth it, imo.
But it's worth it to me to be answerable to no one. I can't, say, sign in to my Google account with an unlocked phone and get banned. Boom, photos gone, no matter how many backups Google made on my behalf. And if I'm gonna do a "takeout" every month and back that up; might as well host myself. Immich is also absolutely amazing software.
I haven't really found a guide or discussions on how best to use Immich in this mirroring way. Admittedly, I haven't looked very hard. Setting up Immich is something that's been on my todo list for years now.
I used to want people to send me full res pics by email to preserve location and find my people easily, but Immich does such a good job with face identification that i stopped worring about what people send me over whatsapp (at least I can find all photos of any one person going back to the first pic i have with them !, my own pics are always geotagged).
Regardless of cloud versions for sharing i need my own local backup. i run immich in docker on Ubuntu, 2 mirrored ZFS drives for redundancy and backed up to rsync.net via restic. (rsync.net i need anyways to back up other data).
Have to try the mobile app still, want to end up with a low res version of my entire library (about 1 TB) on my mobile, face tagged for easy sharing ! Thats next on the list to complete my personal photo management endeavour !
There is, however, a fixed component cost to building such a setup that gets amortized the more disks you buy- I believe it becomes worth it to use a NAS cost-wise over about ~5-6TB of data (and definitively worth it when you get into the tens of terabytes).
Or if you need low-latency storage / any type of performance - neither the cloud nor my homelab has removed my need for my portable NVMe drive
This is also true for businesses. I used to think why do companies don't use open source and save so much on infrastructure costs etc but in reality those costs don't really matter to them and self-hosting costs in terms of time and resources do add up + no gurantee as provided by enterprise level services etc.
There’s nothing as nice as Plex for consuming media IMHO, and home assistant/homebridge glue a pile of cheap IoT stuff into a single, cohesive, useable setup.
I am a reluctant admin. I resent spending time on it. It feels like work, which after 10 hours of doing similar things for much better pay grinds on you. Luckily it doesn’t take much time to keep it going after the initial setup.
Thread and Matter give me hope one day the IoT stuff might be okay. I have no hope I will ever be able to get the single media portal any other way.
And I happily used a Google pro subscription for a while. Even when they doubled the price. Then they started actually limiting it to 5TB and it wasn't worth it.
I guarantee you I am saving boatloads compared to the $100/month tier I'd need for all my data. That pays off my entire hardware budget in under a year and the time I spent is worth a fraction of that.
Even before AI coding I did not trust big tech companies to make updates that have my best interests in mind. Many of these examples are weird feature or ux changes - like how spotify treats liked songs as a playlist (requiring more taps to like a song). Not the end of the world. But on the more extreme side, my gaming PC's windows installation has died via bad windows update twice now. I don't use it for anything important anymore. It's insane that my claude code driven homelab stack has allowed me to have better uptime than github.
Now in the AI coding era, its obvious to me that alot of the tools I am forced to use on a day to day basis are frequently shipped with slop, causing more bugs and all around jankier UX's. With my claude max subscription, im suddenly able to make tools perfectly tailored to what I want, and I can rest assured I wont wake up and find the equivalent of a bad IOS update.
You have to make at least 2 takeout to be sure to have all your files in good shape.
The problem is notto download but to verify and be sure nothing else is missing.
Looking back now it is no big deal. My burden has lessened and i am still the same man I was.
discard/reject/delete: ≈400GB → 25GB → 3GB (< a decade ago; I remember) (I was shocked to see how little of that really mattered).
and since then I have been backing up starting from that 3GB set which, more than a decade later, is obviously more now, but the learning was don't let amount of data control/tame you, it should be the other way round. So I can just use multiple PAYG backup destinations without worrying about insane costs.
The amount of photos and videos I took also went astronomically down after that. I realised what I was missing - I had lost the data and for the most part I didn't have many of those scenes/moments to begin with. Because I was busy taking those digital photos and videos.
unlike Photos, you don’t have a team of professionals managing it 24/7.
(that said, takeout your Google Photos too, because single point of failure, etc).
If you write 2 additional disks per year, this is a $500 initial investment plus $20 per year. Compared to Google's "AI Plus" plan you break even in year 5, assuming Google keeps pricing flat.
Of course you don't get any cloud access or similar.
I do a bit of a hybrid. Borg to expensive, always available storage, but really large stuff that rarely changes and is less important, I exclude and manually throw in glacier once a year or so. Think, all your blue-ray rips, etc.
- Thumbnails and ML vectors consumed hundreds of gigabytes. I had to move the library folder to a dedicated ZFS HDD (700GB currently).
- ImageMagick uses 20GB RAM per Nikon RAW file (this triggered the Linux OOM-killer)
- Container crashes filled the OS drive with 2GB _per_ core dump.
I added `ulimits: core: 0` to the `docker-compose.yml` to stop core dumps and increased VM RAM to 48GB. I also added the following `.env` tweaks:
```
NODE_OPTIONS=--max-old-space-size=8192
VIPS_DISC_THRESHOLD=750m
VIPS_CONCURRENCY=1
MAGICK_MEMORY_LIMIT=8G
MAGICK_MAP_LIMIT=4G
```
```yaml
logging:
driver: "json-file"
options:
max-size: "10m"
max-file: "3"
```
My homelab had spare capacity and I had already offsite backups to Hetzner Storage Box sorted out so it didn't take long or increase my costs.
Very good experience so far. I moved all my photos including multiple complete wedding albums to it, something I couldn't do with Drive without "converting". Nice to finally have all the Google Photos goodies for everything.
Shout out to immich-go, a cli tool for bulk uploads from Google takeout archives.
In hindsight I am baffled that I accepted for google to have all my pictures, scan all the faces, etc. Insane.
But when everyone is doing it, it turns something off in the brain. I attended a seminar where the explanation was: We don’t have the cognitive resource to do a deep analysis (or any analysis) of every single decision every minute, which is why, evolutionary, we relied on crowd wisdom to move such decisions from CPU to ASIC, the ASIC being “is everyone else doing it?”.
I am still angry at myself that I gave them my pictures for years, the pictures of my relatives’ kids, etc.
@Immich team: pleeeeease add the horizon straightening feature.
For the nerds...https://jaisenmathai.com/articles/my-ridiculously-robust-pho...
After using it for a while I actually prefer it to google photos and think it does quite a few things better. I would recommend switching for sure.
I don't think the article actually covers this; they mention installing the immich app on your device and uploading a photo, but it's not at all clear that this will automatically back up all photos taken on your phone.
Photoprism is the best one I’ve found that will just take and index your folders as they are.
Immich will also index your folders as they are. It’s called “External Libraries” in the documentation.
There's https://github.com/alangrainger/immich-public-proxy, which is a weird workaround (it's a separate gallery frontend that syncs with your instance to only expose this specific service, so you can keep your Immich instane private).
I've always considered using something like this but the fact that it's backed up locally (to backup on cloud I have to pay for another service) and not accessible to my family kind of puts me off from setting up something like this
> not accessible to my family
My family has no trouble with it, the UX is excellent. My retirement age parents are regularly adding photos to shared albums and things like that.
I use Apple's top-notch editing for all sorts of stuff - RAW images, videos, basic enhancements, etc. When I tested, Immich didn't have very good, if any, editing features.
I also love Apple's search functions, and it Just Works™. I'm sure Immich has similar, but I know it would take a lot of processing to make it happen.
I also did not really care for Immich's UI, it felt overly basic and Material Design to me, but that is only personal preference.
Another gap is metadata. In Apple Photos I can tap the little "I" icon to see a lot of capture metadata, and in Immich I did not see an equivalent to how good Apple's one is.
It’s less than you think! Consider that Apple’s image recognition happens on your phone. Also, it doesn’t matter if indexing is slow, as long as the photos are indexed before you need to search for them.
I used to run Immich on a 2nd gen Core i3.
I feel like Futo is going in the wrong direction. In the age of AI, anyone can write a DB query, anyone can design a simple UI, and with docker the structure of hosting is no longer a barrier.
I think Futo and Immich should be focusing on data itself. Export, import, store, parse metadata, process, suggest relationships
The sane approach is: Is there something out there with a solid community that fulfils this need? If yes, let’s not vibe code project N.
I think there is this huge wave of projects which will die off when the devs realise the time it takes to maintain it, even with AI.
The amount of code produced will remain significantly higher than before AI, but it will go down a bit, compared to the peak (and I’m not sure we’ve reached peak)
I also have a zfs based fileserver setup were i do snapshots before i backup from my desktop were i manage my pictures.
I backup my zfs based filesystem online for off-side.
My images (a lot of pictures, dslr quality, raw) is around 2tb.
You don't just replicate what a company like Google does for you for 10$ and without knowing stuff about...
I just use Shotwell: https://shotwell-project.org/
For someone like me, it's easy and quick and I have a very robust incremental backup system with air-gapped copies. So it does save me money, because it only takes my time and knowledge.
There is no absolute answer, because it's subjective depending on the needs of the user.
I wouldn't make the instance itself public, but I think this is a decent tradeoff. Of course still requires one to be a sysadmin.
If the company already has a developer team and the software is an expensive domain-specific CRUD app for internal use without significant compliance or risk factors - sure, go DIY.
If the company does not employ any devs, or if the core logic of the software is a business secret of the vendor, or it is customer facing, or there are significant compliance / risk factors - you are better off paying for it.
Unless the cost is astronomical, or the company is motivated also by anger at rent-seeking MBA f*ckers at the vendor.
Tip: export the takeout to Google drive then download it from there.
Otherwise link downloads have # of times you can retry before they are deleted.
And yes Google had gotten progressively every year that I have been a paying customer as well so I am not against your point either. All I am saying is sometimes cloud makes sense financially as well even though open source allows you to do it for free technically. Enshiftication is the word of our age for everything imo unfortunately
For example, paperless-ngx was already an absolute ball of mud, and now there’s a new major version that’s somehow even worse.
I’m just not upgrading. It’s behind a VPN, I control what files it can ingest, so it’s okay if it’s old.
I’m looking at a few alternatives and I’ll switch when they are mature and when I have the time, but there’s essentially no deadline.
(As for data, the third hard disk isn't actually bought yet and is meant to represent an offsite backup. I might use something like AWS Glacier or Hetzner Storage Box for this anyways cuz this is mostly for experimentation. I just like the idea of having a homelab.)
Or you can SSH in and type the key once per reboot.
Codex reviewed my immich setup and recommended I change things the clip model, came up with a better backup strat and found holes in my system.
My codex will ADB into my NVIDIA Shield and proxmox and setup accounts, install packages, create Home Assistant integrations to tie it all together. As a hobbyist, these are things that I would never even think about taking on.
I bought a hdmi-cec controller for my projector last week. Firmware didnt work with the PX3-pro, but it was tied to my ESPHome. Codex diagnosed the issue, rewrote, tested and uploaded the firmware OTA. All done from my phone while I was waiting in line at a store.
I would call it more of a set of tradeoffs
> a lot of the time is spent on doing menial system admin tasks
This is the refreshening opportunity for us non-admins, and a learning opportunity to automate as much of it as possible.
Other benefits not mentioned include not signing your data or life away to whatever Google's lawyers or marketers have decided is best for them, contributing to FLOSS even if only as an end user, not being a victim of rug-pulls...the list goes on.
Obviously this isn't for everybody, because there's no denying it takes time. The point is that for many, this is a very easy tradeoff to make when factoring in all of the other benefits.
This is mostly wording. It is a commitment, it has an opportunity cost, there are tradeoffs.
> This is the refreshening opportunity for us non-admins, and a learning opportunity to automate as much of it as possible.
So many opportunities but the day is still 24 hours long. You still have to choose your opportunities wisely, and give up something else in the process. And if you look at the big picture those who choose to take this opportunity are not so many compared to those who just pay with data so they don't have to pay with time and effort. Everyone pays money in the end.
It's odd that you've gone off on this as opposed to people who spend their 'opportunity time' on Netflix, Call of Duty or D&D.
Sysadmin is a hobby I enjoy and I don't consider it lost time, or wasted opportunity, when so many of my colleagues go out getting drunk or watch football.
"Gone off on this"? Wait till you see me go off on people with a lack of reading comprehension, a pathological persecution complex, and a giant chip on their shoulder.
Anyway, if you ask your parents to explain my comment to you, they'll probably immediately tell you that it's a generic comment on the inability to take up every opportunity all the time. You choose one thing, others choose another, everyone does their own math (you should probably ask someone to do your math for you). I don't care about your hobby which is why I didn't mention "sysadmin", "Netflix", "Call of Duty" or "D&D", nor did I say which one I consider "better" or "worthwhile". Just that you can't do them all.
Right, but the distinction is relevant in the context of the original claim:
> People setting up homelab often talk about money saved but never account for the time they have to spend
My point was that many of us do account for this time in the aforementioned ways.
Plus, once it's set up, it practically runs itself if you have a sane backup methodology and are only running upgrades every couple months.
> jump through this set of hoops
And a weak jab that was for someone who's overwhelmed from trying to parse a couple of sentences. If you thought it's objectively wrong you would have already pointed that out loudly and offensively.
So you just took the opportunity to feel like a big boy and tell me "this admin stuff is not for you". Good for you big boy, glad you got that out of the system.
Of course, you also outsource the effort for almost everything around you (even your reading comprehension), to someone else - an expert, the person you're talking to, or your parents. But you can download a compose file and import it in what's probably a NAS with a nice "import" GUI so welcome to the "target audience". I'm a big boy too, you see, one with actual reading comprehension ;).