Google's Pixel 4 Astrophotography mode is now available on Pixel 2, 3 and 3a devices

Are you certain? Long exposure, small sensor, noise...

@jwasturias I have a Moto G7 - what app is that exactly that you used, can't find it? Sounds good.

@Dave C 150:
I have a G7 Plus. Which has a different camera. 16mp phase detect optical stabilized
.
For the normal G7 try:
https://forum.xda-developers.com/moto-g7/themes/gcam-moto-g7-portrait-night-sight-video-t3906296

For my g7 plus I used this: https://www.getdroidtips.com/google-camera-for-moto-g7-plus/
Strangely, only the complex stuff works e.g. hdr+ night mode etc. A single exposure sometimes does not react at all on the button. But for that just use the native camera. hdr and night mode are worth it. (Considering it is free :) )

Apparently there are lots of ported versions and perhaps some work better than others. Good Luck.

I'm talking about the night sky photography. We all understand that you can select night sight whenever you want

Apologies, I should've been more clear.

RegiStax?

For focus stacking - macrostack/hugin/enfuse. Free and open source.

Sorry, yes, of course software exists but it's like $50 for a single specialized function or some unmaintained open source project. I'm sure there is brilliant software, but I think it's not very efficient.
I think it's kind of an inefficiency of the market that we don't see these features as standard in the bigger photo software suites. Or maybe I'm underestimating the complexity.
Ideal would be a "robust" algorithm that always produces acceptable results on full auto but also allows some manual tweaking.

(Cont) And remember, for a lot of this stuff there is little “artistic interpretation” involved - the end goal is primarily to compensate for hardware limitations. Less noise, better dynamic range, and depth-of-field information (from which background blur can be simulated). There is likely a small set of “optimal solutions” for a given photographic scene.

That, in my mind, is why we need an end-to-end application for computational photography.

...Alright, I tried my backup phone and realize now that this is probably pretty far from Gcam. My new-ish phone is out for repairs so I tested "manually bursted"(because the phone would not save DNG in true ~10fps burst mode) ~1fps shots from the 2015 S6E handheld.
Again 8 frames but this time base ISO, -2ev and testing for DR with a flashlight. This as expected does nothing to raise highlight DR(only my -2ev adjustment made a difference), but also does hardly a thing for shadow DR, which is unexpected. Maybe the sensor in question has too high a noise floor, maybe the DNG format or the specific DNG output from this phone isn't as well supported as ARW, I don't know, but it's just the highlight illuminated by the flashlight that's cleaner, and the surrounding shadows seem no different at all, that's including the corner ~25% that for some reason has a severe purple cast(similar to my new-ish phone though), as well as the center ~75% of the image that at least looks barely useable .

Look up star stacker software. There are several options.

" Sorry, yes, of course software exists but it's like $50 for a single specialized function or some unmaintained open source project. " -DejayRezme

So you want polished software with a slick UI for what, $15? Free? Smartphone companies are sinking a ton of R&D money into automating these things, you can argue that with Google's Pixels you're paying more for their software (as far as the camera tricks and the timely updates) than for the hardware which isn't particularly impressive or special compared to other phones (and I say that as a Pixel user).

If you were complaining about $300+ software licenses I could kinda see the point, but $50 is a bridge too far? Seriously? For something you'll use with kilobuck cameras and lenses? /smh I even mentioned Affinity could do some of these things and that's not a specialized package but a full fledged editor, Enfuse had a lot of different uses too.

(continued)

Maybe if more people actually supported those smaller outfits with their money tho those developers would be able to improve their wares, it's pretty pointless to complain about Adobe then not try any alternatives. Painting open source projects as if they're all chronically abandoned is pretty unfair too. Now if you just wanna slam Adobe for their laziness then I've got no issues with that.

Smartphones' algorithms have control over how the original burst is shot in order to fit the processing tho, so that alone makes it much simpler to automate. There's very different approaches even amongst phones, for hand held shots Google tends to stack identical exposures and priorities zero shutter lag (by having optimized the buffer to reach back in time), other manufacturers use multiple exposures, etc.

Plus as Ottok said above, the sensor readout speed and burst rate are huge factors still in phone's favor. That's all before we get into the really complex part that you *can't* replicate so easily on a PC which is the tile slicing of each frame pre-stack in order to compensate for motion, etc. (that's not really relevant for this astro scenario tho where the camera is expected to be on a tripod and there's no motion in the scene beyond the Earth's rotation)

Sorry, pardon my ignorance. I didn't mean to offend anyone or the software developers of those open source apps.

But with something like a 6fps burst you definitely can reconstruct the motion of the camera optically through computer vision.

“ That's all before we get into the really complex part that you *can't* replicate so easily on a PC which is the tile slicing of each frame pre-stack in order to compensate for motion, ”

I’m not sure I’m following you. Why would a more powerful PC be challenged by this if a mobile computer (smartphone) isn’t?

" But with something like a 6fps burst you definitely can reconstruct the motion of the camera optically through computer vision. " -DejayRezme

To a certain extent, sure, I do handheld HDR and pano stitches with my ILC just fine... However, the phones that are doing the most clever stacking/computational tricks are pulling shots from a constant buffer that's been run at well over 30fps, and that helps a lot with the kinda processing they're trying to do. There can be quite a bit more motion occurring over 2s than over 1/30th of a sec.

" I’m not sure I’m following you. Why would a more powerful PC be challenged by this if a mobile computer (smartphone) isn’t? " -desertsp

It's not about computing power, it's about having a fast enough burst (see above) where you're comparing and averaging almost pixel level differences and which makes it easier to compensate for subject motion, and about the complexity of the algorithms and the R&D thrown behind them.

The camera and/or subject can and likely will move a heck of a lot more over the span of 2s (the time it would take an ILC shot at 6fps to grab a dozen frames) than over a fraction of that with a 5x or faster burst. Now there's a couple of cameras that can now shoot full res/bit-depth RAWs at 30fps or faster, but it's literally just a couple and they only came out within the last 3-4 years.

Google's been iterating on their standard stacking/HDR mode for 6 years now and they were working towards having phones with optimized buffers/hardware for this even before that.

(continued)

As far as the algorithms, what I was alluding to is there's things you can automate fairly easily, like astro stacking or normal HDR, you're literally just lining one frame directly on top of each other, you could even do that completely by hand even if it's more tedious. You can't really replicate what Google/Apple are doing by hand, it would be mind numbing to manually split each frame into dozens of sections and compare each section stack for differences, then re-assemble it etc.

(continued)

Those algorithms can easily be implemented on a PC, stuff like Kandao purports to do exactly that (I've not tried it mind you)... But without an optimized capture method with a fast enough burst, optimal selection of exposures, intimate knowledge about the sensor/lens, etc then it's just a small part of what phones are doing. I'm not saying we'll never get that kinda processing ability in post, heck Google's Pixels can spit out pre-stacked DNGs already; it's just gonna take more work on both ends, not just the software.

Ah I see what you’re saying. Another factor is that stacking in the device only requires storing the final stacked output, versus having to store potentially dozens of images or hundreds of megabytes. Unless of course compression is used, maybe the device stacks the unambiguous parts of the image, but stores individual exposures for trickier areas so the user can determine how to handle them in post.

Phones also have access to gyro data and can align frames based on that and any movement is then treated as subject movement which helps algorithms a lot. Even the tile approach is relatively crude, we've had better segmentation for years, it's just too cpu intensive and even tiles produce acceptable results. In due time we'll have even better segmentation and reconstruction so you will benefit from several frames of arbitrarily moving subject.