Enhance 224 to 176.
Enhance.
Stop.
Move in.
Stop.
Pull out, track right.
Stop.
Center and Pull back.
Stop.
Track 45 right. Stop.
Center and stop.
Enhance 34 to 36.
Pan right and Pull back.
Stop.
Enhance 34 to 46.
Pull back.
Wait a minute.
Go right.
Stop.
Enhance 57 to 19.
Track 45 left.
Stop.
Enhance 15 to 23.
Gimme a hard copy right there.
What is happening on the Photos team? I know a lot of the core team has left already, but lately it really seems like they are cramming in AI functionality in a way that does not consider the whole experience and it truly feels bolted-on. First they tried to mess with regular search and had to leave in a setting for people to go out of their way to disable it.
Google Photos used to be one of my most favorite apps. Now it's some testing playground. I'm sure some new startup is approaching this area so the cycle can repeat.
This is all of Google ever since the "Red Alert" when ChatGPT came on the scene. It just took a glacial 3 years for it to come to fruition like everything at Google.
For real, even Lowes is doing this crap. I tried looking up a some products on their website with my phone the other day and every product page is now partially obscured by a useless AI widget that can't be removed. They've made their product categorically worse just to check the AI box.
Remember when FB really took off, and places like Lowes tried to make their site be a "social hub" too? Like you'd add your friends there and follow their shoppings or whatever
> but lately it really seems like they are cramming in AI functionality in a way that does not consider the whole experience and it truly feels bolted-on.
I feel that everywhere, not just Google. Windows is probably one of the most egregious with copilot being jammed into every app.
None of this "AI" stuff feels integrated or even thought out at all. The whole thing is just bolted on and it feels like the only reason is so they can say "see, we have AI too! Now give us money" to investors.
And they took away the one editing feature I use the most -- the perspective correction tool. I take a lot of photos of documents, and to avoid shadows I often can't take the photo from above, so used to use the perspective tool to square up the document.
I use zero of the new AI tools, but they took away the one tool I really want.
They claim they'll bring it back...sometime.
"Perspective tools – Our team is working to restore these"
Lately they also shrunk the image preview, now it's like 60% of the phone screen estate - good look hunting for details while editing. I don't think, change my mind, that they care about real users feedback.
Managers on AI projects need to show "mass adoption" in order to survive the next annual review. That's why you have pointless AI stuffed into every orifice of Google/YouTube.
One of the reasons why I think Google Photos is pushing out this functionality is that these new AI-generated variations count toward your overall storage.
So hey, you take a selfie, 3 MB. And now you want your selfie to show yourself posing in front of the Eiffel Tower, the Statue of Liberty, the Great Wall of China, and the Grand Canyon. OK, now you've added an additional 12 MB.
Do that often enough, and you'll have to buy another block of Google One storage, which is MRR for Google.
Of course they don't. But if even a small percentage of their userbase uses this on even a small percentage of their photos, that amounts to a massive amount of storage. It doesn't cost Google very much to store all those images, but those users will have to pay for more storage. Even thought it isn't significant to Google when a single person does it, it becomes highly significant when millions of people do it.
A million people adding 12MB each would be 12TB in increased storage sales. Is that really significant in increased profit if you set it off against the cost of providing the storage, the computational cost of generating 12TB of images, plus the salaries of the developers to write and then maintain this feature?
I think it's more likely that they're doing it because they think it makes their product better, increasing sales that way instead.
This line of reasoning doesn’t make sense because 3MB is marginal for the amount of work put in to develop this feature.
Google photos offers a save as copy / save as original feature when editing photos and videos. Removing the save as original button would be cheaper and significantly more effective.
Nope - its so they can sell you ads - it learns about your interests based on your prompts. That marginally increased storage space won't make much of a dent.
The biggest function I'm missing is the keystone transformation, but magic eraser is now a cloud-only function that is much worse, they have rearranged the controls so in a way that separated some of the color corrections, and the resize edit handles for the functions that are still available are now rounded in a way that makes it much less clear what you are doing.
Same here for all of the 'self hosted' things. I've also just realized that my UDM has a wireguard implementation titled 'teleport' that seems to do basically the same thing. This is nice because sometimes the server at home that is hosting tailscale breaks, and the UDM will let me into the network at the router level. Knock on wood, that's yet to crash.
I host Immich in Hetzner (VPS w/ attached 1TB storage box) and connect to it using Tailscale which works pretty flawlessly on my phone. It's great, although the VPS is pretty slow and I might move to a home server at some point.
I might also just switch over to Ente so I don't have to deal with the self-hosting. Price for Ente is about equivalent for what I'm paying Hetzner right now.
I’m personally wary of uploading too much private data to any host. I am also a customer of Hetzner, and rent a couple of bare metal servers. But I mostly use it to store data that it wouldn’t be that big of a deal if it was stolen by someone.
I’ve previously experimented a bit with encrypted volumes that I manually decrypt over ssh, and even full disk encryption that I manually decrypt over ssh.
My experience with Hetzner has been good. It is really rare that the servers go down on their own. Reboots are usually my own doing, so I am already “around” to decrypt encrypted volumes.
I have experienced critical, unrecoverable hardware failure on Hetzner servers a couple of times over the years. But I’ve had offsite backups in place since day one, so I never ultimately lost any important data. Had to deprovision the broken server, reprovision a new one and restore from my offsite backup. Which is a bit of a hassle, but no biggie because the only one that relies on my servers is mostly myself. A few days of downtime because I am too busy to set up a new server right away is therefore also ok for me, with how infrequently it has happened.
A single Hetzner server should never be the only place hosting a copy of all your photos or other data you cannot afford to lose. But that applies to any host really. Not unique to Hetzner.
> A single Hetzner server should never be the only place hosting a copy of all your photos
Hetzner (or any vps provider) should not be a place at all to store ANY copy of your photos, unencrypted.
I agree that they respect privacy a lot, they're probably the best of all the service providers when it comes to your data and that there are data protection laws in place etc etc
but in the end, it's your personal photos, I wouldn't be willing to upload it to any provider unencrypted, good that you're encrypting
I took care of it myself by cancelling my rental of the server and renting a different one. And then setting up that one the way I wanted it and restoring data from my offsite backup.
I think there was a form asking for reason for cancelling the server and I ticked something like “other” and left a note for them saying that there was hardware problems. So I would assume they have a look at it, replace the bad components and then rent it out to someone else.
Huh. Was curious what kind of discount they provide in case of downtime caused by hardware failure, but it sounds like they didn't even notice. Shouldn't they monitor the basic vitals? Or you simply reacted too quickly?
You could keep the Hetzner VPS with storage for faster online serving of assets and connect a second immich instance only for machine learning on your home server. That way you'd get the best of both worlds: fast media serving and higher performance. That would mean that images are uploaded to the Hetzner server, but the compute-intensive image classification takes place on your home server.
Tailscale works great, but it's annoying to have to have an always on VPN on android for it. I need to switch tailscale off if I need to switch to another VPN.
Also when I have it on my private DNS stops working, which to be fair I haven't put a huge amount of effort into solving yet.
I love it for things like ssh to a server at home, but for things like hosting a service I prefer something like cloudflare tunnel or a self hosted reverse proxy. Though tailscale funnel looks promising.
I seriously dislike adding a package source for a single application. It feels dirty to me. I can't explain it but it makes me feel like I need to take a shower.
I don't use arch but this looks cleaner than whatever Debian or fedora (both of which I use) have going on
Alternatively, I feel much better when the upstream vendor is the one packaging and signing the software I install, instead of a (possibly malicious) volunteer from my distro's repository team.
I use PhotoSync to transfer files automatically. There is also icloudpd and other tools to grab them from iCloud but it requires Advanced Data Protection off.
I was just thinking something like this would be nice. I wanted to remove a shadow from a picture but lasso tools weren't picking it up. After some Googling found a tool in Samsung Gallery to do it. It still tried to steer me towards generative edits instead though.
But just telling it "remove the shadow from this image" would be great.
Glad to see Ente getting some love here. Immich gets a lot of the attention, but Ente also allows you to self host as an exit ramp if their pricing/antics ever get out of hand.
Yeah, I'm alright. They're really desperately trying to find something that gets people excited about all the AI garbage they're shoveling at us. It's sad. Maybe find something that actually solves a problem instead of just sunk-cost-fallacying LLM crap.
Same, the google phones just absolutely hated that I went into airplane mode frequently but still wanted to do things like listen to music, view and edit notes, etc. God forbid I try to add to my shopping list while camping.
One compromise in GrapheneOS is to use Storage Scopes to limit the scope of Google Photos to select folders, or potentially even disable Network permissions for it altogether. I use Google Photos for sharing photos with my partner and family, and I can do this by moving photos around on disk so that they enter the scope. The camera app from Google is great, and so is the keyboard: neither gets Network permission on my phone.
I'm on a Pixel right now as well and, just like you say, it's the software experience that sells it. Enough to justify my next phone also being a Pixel? Jury is still out on that.
I turn a lot of real photos into cartoons. I love the feature!
Most recently I took a photo of my grandma and me, asked Gemini to make it a cartoon, asked Gemini to make the new variant into a birthday card.
My grandma loved it! I was happy to make her something custom. Buying people cards just never felt right to me. Writing was also never my strongest suit - so this new form of expression for me has been enhancing :)
The only remaining thing I need to do better is getting the card printed! I wish that also was only 12seconds of work.
Have you never taken a time-sensitive photo that was otherwise perfect but has an annoying visual obstruction (e.g. traffic sign poles, a reflection on a windshield for photos taken out of a moving car etc.)?
People don't just use cameras as perfect recording tools, as wild as that sounds. Photos are also for expression. And so, editing can be used to enhance the expression part, getting the image away from the recording parts maybe, but closer to the original intention of taking that photo.
How exactly does it capture the moment? Almost certainly, the camera doesn't capture what your eyes do, even without advanced computational photography because of lens effects, color range, etc...so what moment is being captured? The moment of the camera?
(this is from the original Blade Runner, for anyone wondering)
I love this being quoted here because while it superficially seems that this interaction is interacting with an AI in a conversational way, it shows tool use that is at the opposite end of what interacting with modern AI tends to be like/what AI marketing wants it to be.
In this scene:
- the user is guiding the tool step by step to a desired outcome, rather than giving a broad vague result that the tool gets to arbitrarily
- the user knows intimately the specific technical capabilities of the tool
- every command is phrased to be unambiguous and lead to a deterministic result
Imagine if instead all of the above, Deckard just gave a broad request, eg “can you help me find who the killer is by analyzing this picture”?
Okay that's a cool feature. This is a good place to use this technology.
It is interesting, though, that Google's products have suffered so much regression as AI has advanced. In the past I could search my images with "Subaru" and reliably get photos. Now it's clearly a subset.
Same with Google's voice assistant which was actually more capable 5 or 6 years ago.
I get it. It's probably easier to maintain with a general purpose LLM or diffusion model behind the scenes or whatever. It's just a pity that the PMs and engineers who made the good stuff have either lost their positions or have changed their minds.
Imagine one day that Tesla will only allow you to control your car with voice commands. It would be hilarious. Is conversation UX suitable for everything? I think not.
Turn left, no other left, no the left after that left. No, damn it stop the car.
...but why not by voice? "Vibe editing" images by voice doesn't offend my 20+ years of developing Photoshop skills the way that typing-out an imagined conversation between MacGyver and a 1980s image-editing-computer does. Oddly, I can't explain why either.
----
What I dislike the most about "prompts" being the default input for AI models thesedays is the inability to "browse" its featureset to see what it's actually capable of. I don't want to spend minutes/hours throwing different natural-language commands at it to seeing if it understands chroma-keying from chromatic-aberration - or if it can do lossless JPEG block-level transformations. It's when something stops being a useful tool but a hinderance or even a toy (perhaps even with Achievements and microtransaction unlockables).
As usual the comments here are very negative on anything and everything AI. This will definitely have appeal for normal users outside of HN bubble. This is also why Google is in a unique position to be able to really capitalize on AI: they already have users that they can ship to vs the next YC startup being able to hit critical mass.
If you want to try well-engineered neural network use, you should be trying Adobe's products. They have integrated these features with far more tact and actual benefit than I would've expected from them. Google is embarrassing themselves by stuffing AI in places it doesn't fit, and Microsoft is worse.
You're joking right? Google is integrating photo editing features that can exist at the point of inception. It doesn't get more integrated or fit better than that.
But it shouldn't have been conceived with that design. Adobe products blend a timesaving tool into a truly productive workflow. Photoshop, a layer or a filter. Lightroom, a brush. Mystery-meat autocorrect/enhance buttons only get you so far and may alter far more than you want, which is dangerous when you want to want to present a slightly polished version of reality.
Just yesterday I wanted to simply take the first frame of one of those stupid "motion" photos (that someone else took, but could have happened to me given the weekly UI churn). I tried really hard, but it's just not possible.
What I did find was many, many different buttons and icons that start some kind of AI magic enhancement crap.
I look forward to trying:
Enhance 224 to 176. Enhance. Stop. Move in. Stop. Pull out, track right. Stop. Center and Pull back. Stop. Track 45 right. Stop. Center and stop. Enhance 34 to 36. Pan right and Pull back. Stop. Enhance 34 to 46. Pull back. Wait a minute. Go right. Stop. Enhance 57 to 19. Track 45 left. Stop. Enhance 15 to 23. Gimme a hard copy right there.
What is happening on the Photos team? I know a lot of the core team has left already, but lately it really seems like they are cramming in AI functionality in a way that does not consider the whole experience and it truly feels bolted-on. First they tried to mess with regular search and had to leave in a setting for people to go out of their way to disable it.
Google Photos used to be one of my most favorite apps. Now it's some testing playground. I'm sure some new startup is approaching this area so the cycle can repeat.
This is all of Google ever since the "Red Alert" when ChatGPT came on the scene. It just took a glacial 3 years for it to come to fruition like everything at Google.
> lately it really seems like they are cramming in AI functionality in a way that does not consider the whole experience and it truly feels bolted-on.
"They" could be just about any major software vendor as of late, it's hardly specific to Google.
Apple is avoiding it more somewhat, but perhaps more due to being behind on the tech itself
Are they? I just took iPadOS 26 and it wouldn't shut up about all the new AI stuff
For real, even Lowes is doing this crap. I tried looking up a some products on their website with my phone the other day and every product page is now partially obscured by a useless AI widget that can't be removed. They've made their product categorically worse just to check the AI box.
Remember when FB really took off, and places like Lowes tried to make their site be a "social hub" too? Like you'd add your friends there and follow their shoppings or whatever
> but lately it really seems like they are cramming in AI functionality in a way that does not consider the whole experience and it truly feels bolted-on.
I feel that everywhere, not just Google. Windows is probably one of the most egregious with copilot being jammed into every app.
None of this "AI" stuff feels integrated or even thought out at all. The whole thing is just bolted on and it feels like the only reason is so they can say "see, we have AI too! Now give us money" to investors.
> Windows is probably one of the most egregious with copilot being jammed into every app.
Seeing copilot shoved into Notepad of all things was probably the worst
And they took away the one editing feature I use the most -- the perspective correction tool. I take a lot of photos of documents, and to avoid shadows I often can't take the photo from above, so used to use the perspective tool to square up the document.
I use zero of the new AI tools, but they took away the one tool I really want.
They claim they'll bring it back...sometime.
"Perspective tools – Our team is working to restore these"
> Google Photos used to be one of my most favorite apps. Now it's some testing playground.
we have entered a new epoch: The Ensloppocene
Lately they also shrunk the image preview, now it's like 60% of the phone screen estate - good look hunting for details while editing. I don't think, change my mind, that they care about real users feedback.
Managers on AI projects need to show "mass adoption" in order to survive the next annual review. That's why you have pointless AI stuffed into every orifice of Google/YouTube.
One of the reasons why I think Google Photos is pushing out this functionality is that these new AI-generated variations count toward your overall storage.
So hey, you take a selfie, 3 MB. And now you want your selfie to show yourself posing in front of the Eiffel Tower, the Statue of Liberty, the Great Wall of China, and the Grand Canyon. OK, now you've added an additional 12 MB.
Do that often enough, and you'll have to buy another block of Google One storage, which is MRR for Google.
Sure, the company of the "I don’t know how to count that low" meme cares about 12MB.
Of course they don't. But if even a small percentage of their userbase uses this on even a small percentage of their photos, that amounts to a massive amount of storage. It doesn't cost Google very much to store all those images, but those users will have to pay for more storage. Even thought it isn't significant to Google when a single person does it, it becomes highly significant when millions of people do it.
A million people adding 12MB each would be 12TB in increased storage sales. Is that really significant in increased profit if you set it off against the cost of providing the storage, the computational cost of generating 12TB of images, plus the salaries of the developers to write and then maintain this feature?
I think it's more likely that they're doing it because they think it makes their product better, increasing sales that way instead.
This line of reasoning doesn’t make sense because 3MB is marginal for the amount of work put in to develop this feature.
Google photos offers a save as copy / save as original feature when editing photos and videos. Removing the save as original button would be cheaper and significantly more effective.
I upgraded to 2TB of storage since 15GB just wasn’t cutting it. Two years later, I’m already at a whopping 200GB!
similar on the 2TB, but at 500GB. Having toddlers will do that. I'm trying to capture so much.
Same here. 10mo old. Funny thing is I have barely revisited the photos and videos, due to lack of time.
Nope - its so they can sell you ads - it learns about your interests based on your prompts. That marginally increased storage space won't make much of a dent.
It's possible for two things to be true at the same time
Meanwhile they nuked a bunch of actually useful conventional and device-local editing functionality in the same app.
Oh no. What did they remove?
The biggest function I'm missing is the keystone transformation, but magic eraser is now a cloud-only function that is much worse, they have rearranged the controls so in a way that separated some of the color corrections, and the resize edit handles for the functions that are still available are now rounded in a way that makes it much less clear what you are doing.
The crop interface got way worse in the latest update (the "comic sans" update) as well.
Yeah, I don't understand how that update got pushed by anyone who actually used the product. Like, they broke some of the most used functions.
What's the comic sans update?
The latest android update where they made all the fonts cartoony, added weird decorations to the volume slider, and messed with the camera edit menus.
Oh, I had a suspicion that was what you meant, that's a good name for that annoying change.
I think it's Android 16 QPR1 (the September update), it changed the font on a bunch of UI text to something rounded and perhaps cartoonish.
Not affiliated, but Immich (https://immich.app/) is a great self-hosted alternative to Google Photos.
Unfortunately it seems to still be lacking HDR images support ( https://github.com/immich-app/immich/discussions/7262 ) if you care about that or not.
I really tried my best with Immich, but because I didn't want to open a port on my home firewall it made syncing kind of a pain.
I ended up going with Ente and have been pretty happy with it.
Tailscale is what I use to solve such issues.
Same here for all of the 'self hosted' things. I've also just realized that my UDM has a wireguard implementation titled 'teleport' that seems to do basically the same thing. This is nice because sometimes the server at home that is hosting tailscale breaks, and the UDM will let me into the network at the router level. Knock on wood, that's yet to crash.
Wireguard is the answer. That's how I use it
Interesting. Not OP, but I'm looking for something that can punch through corporate firewalls so I can use this (and other) software at work.
ZeroTier? https://www.zerotier.com/
Oh my god, the syncing requires opening a port??? The amount of hoops these applications require us to jump through nowadays
I host Immich in Hetzner (VPS w/ attached 1TB storage box) and connect to it using Tailscale which works pretty flawlessly on my phone. It's great, although the VPS is pretty slow and I might move to a home server at some point.
I might also just switch over to Ente so I don't have to deal with the self-hosting. Price for Ente is about equivalent for what I'm paying Hetzner right now.
Immich didn't have encryption last time i checked, do you trust hetzner with your photos?
I also use Immich, but on a local server (using tailscale to reach it from outside)
I’m personally wary of uploading too much private data to any host. I am also a customer of Hetzner, and rent a couple of bare metal servers. But I mostly use it to store data that it wouldn’t be that big of a deal if it was stolen by someone.
I’ve previously experimented a bit with encrypted volumes that I manually decrypt over ssh, and even full disk encryption that I manually decrypt over ssh.
My experience with Hetzner has been good. It is really rare that the servers go down on their own. Reboots are usually my own doing, so I am already “around” to decrypt encrypted volumes.
I have experienced critical, unrecoverable hardware failure on Hetzner servers a couple of times over the years. But I’ve had offsite backups in place since day one, so I never ultimately lost any important data. Had to deprovision the broken server, reprovision a new one and restore from my offsite backup. Which is a bit of a hassle, but no biggie because the only one that relies on my servers is mostly myself. A few days of downtime because I am too busy to set up a new server right away is therefore also ok for me, with how infrequently it has happened.
A single Hetzner server should never be the only place hosting a copy of all your photos or other data you cannot afford to lose. But that applies to any host really. Not unique to Hetzner.
> A single Hetzner server should never be the only place hosting a copy of all your photos
Hetzner (or any vps provider) should not be a place at all to store ANY copy of your photos, unencrypted.
I agree that they respect privacy a lot, they're probably the best of all the service providers when it comes to your data and that there are data protection laws in place etc etc
but in the end, it's your personal photos, I wouldn't be willing to upload it to any provider unencrypted, good that you're encrypting
Also, check this out (not my project): https://github.com/rfjakob/gocryptfs
> I have experienced critical, unrecoverable hardware failure on Hetzner servers a couple of times over the years
How do they handle such situations?
I took care of it myself by cancelling my rental of the server and renting a different one. And then setting up that one the way I wanted it and restoring data from my offsite backup.
I think there was a form asking for reason for cancelling the server and I ticked something like “other” and left a note for them saying that there was hardware problems. So I would assume they have a look at it, replace the bad components and then rent it out to someone else.
Huh. Was curious what kind of discount they provide in case of downtime caused by hardware failure, but it sounds like they didn't even notice. Shouldn't they monitor the basic vitals? Or you simply reacted too quickly?
You could keep the Hetzner VPS with storage for faster online serving of assets and connect a second immich instance only for machine learning on your home server. That way you'd get the best of both worlds: fast media serving and higher performance. That would mean that images are uploaded to the Hetzner server, but the compute-intensive image classification takes place on your home server.
Tailscale solves the open port thing for me
Tailscale works great, but it's annoying to have to have an always on VPN on android for it. I need to switch tailscale off if I need to switch to another VPN.
Also when I have it on my private DNS stops working, which to be fair I haven't put a huge amount of effort into solving yet.
I love it for things like ssh to a server at home, but for things like hosting a service I prefer something like cloudflare tunnel or a self hosted reverse proxy. Though tailscale funnel looks promising.
I seriously dislike adding a package source for a single application. It feels dirty to me. I can't explain it but it makes me feel like I need to take a shower.
I don't use arch but this looks cleaner than whatever Debian or fedora (both of which I use) have going on
https://tailscale.com/kb/1036/install-arch
Alternatively, I feel much better when the upstream vendor is the one packaging and signing the software I install, instead of a (possibly malicious) volunteer from my distro's repository team.
With a scary header about not relying on it as your sole photo library.
...because it's still in beta, they're aiming for a first stable release this year.
It's a good idea to have backups/multiple copies regardless of what system you're using.
Interesting. any alternative options for iCloud Photos?
You can import your iCloud photos into Immich. Immich itself is agnostic to the photos and mostly just operates on folders of files.
https://github.com/simulot/immich-go is what I used to import ~300gb of photos from Google Photos to Immich. Not sure how well it works for iCloud.
There is also https://ente.io/ which is a private & secure photo backup app that you don't need to self-host.
> Not sure how well it works for iCloud.
iCloud Photos stores originals (assuming you're using "Download Originals to this Mac") in the "originals" folder of your .photoslibrary package.
If you don't sync your originals, use iCloud Photos Downloader (https://github.com/icloud-photos-downloader/icloud_photos_do...) to get them.
I use PhotoSync to transfer files automatically. There is also icloudpd and other tools to grab them from iCloud but it requires Advanced Data Protection off.
Not affiliated, but Immich (https://immich.app/) is a great self-hosted alternative to iCloud Photos.
Doesn’t do videos, that’s a dealbreaker for me
Of course it does...
I was just thinking something like this would be nice. I wanted to remove a shadow from a picture but lasso tools weren't picking it up. After some Googling found a tool in Samsung Gallery to do it. It still tried to steer me towards generative edits instead though.
But just telling it "remove the shadow from this image" would be great.
I first read that as controversial editing. I'm not sure I was wrong.
Glad to see Ente getting some love here. Immich gets a lot of the attention, but Ente also allows you to self host as an exit ramp if their pricing/antics ever get out of hand.
Idk who talks to a phone out loud in order to edit a picture. Was there a market asking for this function?
Conversational here means "using natural language", not "talking out loud".
Ah, well in that case I can see how a segment of the population would prefer describing the edits.
Yeah, I'm alright. They're really desperately trying to find something that gets people excited about all the AI garbage they're shoveling at us. It's sad. Maybe find something that actually solves a problem instead of just sunk-cost-fallacying LLM crap.
The "AI boom" is really just "sunk cost fallacy At Scale" it's incredible
Today on Gemini one of their suggested activities is to upload a selfie and have it make you look '80s.
I uploaded a selfie with a kinda-sorta '80s haircut and it completely changed my gender.
Does this require associating your phone with a G account?
I believe so. I'd imagine the image has to be uploaded to Google Photos as well, since most of their nice features like personal tagging require that.
Part of the reason I switched to an iPhone was for all the on device stuff that Google insists you use the cloud for.
Same, the google phones just absolutely hated that I went into airplane mode frequently but still wanted to do things like listen to music, view and edit notes, etc. God forbid I try to add to my shopping list while camping.
One compromise in GrapheneOS is to use Storage Scopes to limit the scope of Google Photos to select folders, or potentially even disable Network permissions for it altogether. I use Google Photos for sharing photos with my partner and family, and I can do this by moving photos around on disk so that they enter the scope. The camera app from Google is great, and so is the keyboard: neither gets Network permission on my phone.
So many companies just about to find out that their apps were features not products.
Enhance. Enhance. Enhance.
Honestly surprised to see such a quick turn around on what was a Pixel 10 exclusive feature. I guess they did that for Circle to Search too.
Glad to see that these features aren't artificially limited to specific hardware, often under the pretense that some advanced "AI chip" is necessary.
These features are limited to a specific demographic (those who are 18+ and live in the US) and are 'eligible' (whatever that means) though.
Google certainly has to walk a fine line here because it's not like the hardware they are selling is compelling by itself.
True. Being "stock" (free of unremovable spyware/crapware) and getting OS updates are the primary benefits of Pixel devices for me.
I'm on a Pixel right now as well and, just like you say, it's the software experience that sells it. Enough to justify my next phone also being a Pixel? Jury is still out on that.
Uhm... I'm probably "consevative" but if I take a photo I want to have a photo of the moment and not "a variation of the moment"... o_O
The original is always kept and is easy to see
I turn a lot of real photos into cartoons. I love the feature!
Most recently I took a photo of my grandma and me, asked Gemini to make it a cartoon, asked Gemini to make the new variant into a birthday card.
My grandma loved it! I was happy to make her something custom. Buying people cards just never felt right to me. Writing was also never my strongest suit - so this new form of expression for me has been enhancing :)
The only remaining thing I need to do better is getting the card printed! I wish that also was only 12seconds of work.
That's ok. You can... not use that feature.
Have you never taken a time-sensitive photo that was otherwise perfect but has an annoying visual obstruction (e.g. traffic sign poles, a reflection on a windshield for photos taken out of a moving car etc.)?
No, not relly. If there is a traffic sign post then removing it makes it kinda... "untrue"?
No because it captures the moment.
People don't just use cameras as perfect recording tools, as wild as that sounds. Photos are also for expression. And so, editing can be used to enhance the expression part, getting the image away from the recording parts maybe, but closer to the original intention of taking that photo.
How exactly does it capture the moment? Almost certainly, the camera doesn't capture what your eyes do, even without advanced computational photography because of lens effects, color range, etc...so what moment is being captured? The moment of the camera?
Still no way to sync the "favorites" album between the default Android Photos app and Google Photos, yet they're busy building this slop.
Not to mention the dark patterns that attempt to trick you into backing up your entire photo library, over and over again.
Or the inability to exclude folders from the backup process.
Maybe get the basic expectations of a Photo app right before adding features nobody asked for.
- Enhance 34 to 36.
- Pan right and pull back. Stop.
- Enhance 34 to 46.
- Pull back. Wait a minute. Go right. Stop.
- Enhance 57-19.
- Track 45 left. Stop.
- Enhance 15 to 23.
- Give me a hard copy right there.
(this is from the original Blade Runner, for anyone wondering)
I love this being quoted here because while it superficially seems that this interaction is interacting with an AI in a conversational way, it shows tool use that is at the opposite end of what interacting with modern AI tends to be like/what AI marketing wants it to be.
In this scene:
- the user is guiding the tool step by step to a desired outcome, rather than giving a broad vague result that the tool gets to arbitrarily
- the user knows intimately the specific technical capabilities of the tool
- every command is phrased to be unambiguous and lead to a deterministic result
Imagine if instead all of the above, Deckard just gave a broad request, eg “can you help me find who the killer is by analyzing this picture”?
We'll really be in the future when this feature allows looking around corners like in the movie.
You can already do that with some of the other AI generators. Whatever you see around the corner is all hallucinated of course.
Real-life Enemy of the State, but in the worst possible way.
Okay that's a cool feature. This is a good place to use this technology.
It is interesting, though, that Google's products have suffered so much regression as AI has advanced. In the past I could search my images with "Subaru" and reliably get photos. Now it's clearly a subset.
Same with Google's voice assistant which was actually more capable 5 or 6 years ago.
I get it. It's probably easier to maintain with a general purpose LLM or diffusion model behind the scenes or whatever. It's just a pity that the PMs and engineers who made the good stuff have either lost their positions or have changed their minds.
"in the US" FFS...
Imagine one day that Tesla will only allow you to control your car with voice commands. It would be hilarious. Is conversation UX suitable for everything? I think not.
Turn left, no other left, no the left after that left. No, damn it stop the car.
/s
Conversational via text, not via voice
...but why not by voice? "Vibe editing" images by voice doesn't offend my 20+ years of developing Photoshop skills the way that typing-out an imagined conversation between MacGyver and a 1980s image-editing-computer does. Oddly, I can't explain why either.
----
What I dislike the most about "prompts" being the default input for AI models thesedays is the inability to "browse" its featureset to see what it's actually capable of. I don't want to spend minutes/hours throwing different natural-language commands at it to seeing if it understands chroma-keying from chromatic-aberration - or if it can do lossless JPEG block-level transformations. It's when something stops being a useful tool but a hinderance or even a toy (perhaps even with Achievements and microtransaction unlockables).
As usual the comments here are very negative on anything and everything AI. This will definitely have appeal for normal users outside of HN bubble. This is also why Google is in a unique position to be able to really capitalize on AI: they already have users that they can ship to vs the next YC startup being able to hit critical mass.
If you want to try well-engineered neural network use, you should be trying Adobe's products. They have integrated these features with far more tact and actual benefit than I would've expected from them. Google is embarrassing themselves by stuffing AI in places it doesn't fit, and Microsoft is worse.
You're joking right? Google is integrating photo editing features that can exist at the point of inception. It doesn't get more integrated or fit better than that.
But it shouldn't have been conceived with that design. Adobe products blend a timesaving tool into a truly productive workflow. Photoshop, a layer or a filter. Lightroom, a brush. Mystery-meat autocorrect/enhance buttons only get you so far and may alter far more than you want, which is dangerous when you want to want to present a slightly polished version of reality.
meta is the worst offender
Just yesterday I wanted to simply take the first frame of one of those stupid "motion" photos (that someone else took, but could have happened to me given the weekly UI churn). I tried really hard, but it's just not possible.
What I did find was many, many different buttons and icons that start some kind of AI magic enhancement crap.
Pull up, tap on "shots in this photo", scrub to the left, tap on "save copy". It seems fairly straightforward to me.
Thank you!
But I would never have guessed in my life that there is a "pull up" gesture there.