Hyperlapse, Instagram's new video stabilization app
wired.com
wired.com
PDF here: https://research.microsoft.com/en-us/um/redmond/projects/hyp...
Examples here: http://research.microsoft.com/en-us/um/redmond/projects/hype...
It's even more weird that the WIRED writer didn't mention this. It was major news all over the place two weeks ago. Good PR folks at Instagram / FB.
This is more comparable to if Apple made an application called "Camera" and then Google did the same.
Microsoft did not invent hyperlapse last week. They showed off a method of producing a hyperlapse. Instagram now has another method of producing a hyperlapse.
Here's a video that, coincidentally, was also published last week about shooing a hyperlapse with a DSLR.
It looks like Dan Eckert coined the term for this video: https://vimeo.com/19441262 then later used it to refer to the general technique, for which he registered http://www.hyperlapse.com/ in 2012. Other people started using it, like https://vimeo.com/50238512, and now Microsoft and Facebook are both developing applications that both use the term to automatically create hyperlapse sequences.
This article seems like an exclusive that Wired got, probably been in the works for ages and timed for the release date.
Fine line between this kind of thing and just letting the company write the article themselves (aka "native advertising".)
Turning atomic consumption units into an ad units (actual definition of native ads) is good for users. It creates a seamless experience and puts the onus on advertisers to engage authentically in a native way. It encourages content creation and doesn't interrupt our lives.
Yes, advertorials have a foul taste about them and seem more manipulative than anything else...but its time we stop bashing native ads, advertorials are a very small piece of the native ad world.
And yes, the concept existed well before both.
Note: FB employee.
Microsoft Research was manually computing a motion map using heavy image-processing techniques that they then used to decide which frames to use and how to align them. Their work can do things like throw away a series of 'noisy' frames that wouldn't fit the overall scene motion. For instance, if you had a head mounted camera and were riding around on your bike, it could throw away the moments where you look both ways before crossing the street.
Instagram's technology utilizes extra sensor data to make their frame corrections. As far as I can tell from the brief article, the technology doesn't do much besides stabilization. That's not to belittle it, as it's a really neat application demonstrating the benefits of mobile sensor integration. And you might be able to create a similar motion map using the gyroscope data that the folks at Microsoft create through their image processing techniques!
The gyroscope data most certainly would not allow one to do the same.
The usefulness (or at least coolness) of Hyperlapse is immediately obvious upon using it and it’s a good fit for smartphones, unlike the time lapse feature.
Smartphones are small and light. Standing them on an edge is in the best case precarious and in the worst case impossible (because the edges are sometimes rounded). Depending on where you prop them up (e.g. stone walls) it might also be damaging to the phone. Also, since they are so small and light (or not so much small, just not really ergonomically shaped for steady holding) holding them steady is a real challenge, especially when moving.
Hyperlapse tackles these problems with time lapse on a phone very well. It’s much more thoughtful and considerate of the hardware it’s running on than a simple time lapse feature.
(Plus, that UI is fucking amazing. One button to record, one slider to adjust the speed. This is inside the bounds of complexity Apple lets in their own very basic camera app. The slow motion feature Apple offers has similar UI complexity. So this isn’t even a super-complex feature from a user perspective.)
Huh? There are 200000 mounts available for all kinds of tripods.
If you need to mount your device and buy a mount to actually mount your smartphone then that is and always will be a niche feature. (Also, practically all mounts don’t help a lot when moving the phone around.) It’s not completely useless, but it is useless to most people. Hyperlapse is a much more useful and device-appropriate implementation of the same idea, more or less, that works without the user really having to be very careful or creative or having to buy anything. Plus, the UI is super-simple and not at all confusing.
I first read about this app today, downtown, while drinking a coffee. To get decent video out of it I had to hold the camera in front of me and walk around a bit, not really trying to hold it super-steady, and I got cool results. Also, I didn’t have to buy a mount. Do the same with the time lapse feature and it’s a disaster.
Just because something is possible doesn’t mean it’s a great experience – or a great addition to the OS. My point is this: Hyperlapse is much more useful than the time lapse feature for most users, all the while not being more difficult to use.
What same idea? These do entirely different things. Hyperlapse is a time-sped-up video with intelligent image stabilization, and timelapse is, well, a timelapse.
Not the same idea or feature at all. In fact even Apple themselves describe timelapse for what it is: a timelapse feature for static scenes: "Capture the experience of the sun setting, a city street bustling, or a flower blooming with the new Time-lapse mode in Camera".
Sounds plausible, but the "hence no Android version yet" seems like the standard operating mode for Instagram and most startups. It's not like most come with Android versions from the start.
The price change makes it obvious that he wants to make money off this opportunity.
Apple store is not great on launch day for new hot products - but almost always within 24-36 hours, the proper app gets moved into position #1. (For some definition of proper)
See T&L's online Google Maps Hyperlapse tool, from 2012:
[0] http://en.wiktionary.org/wiki/Wiktionary:Frequency_lists/TV/...
Seriously, it's a shame DSLR etc manufactures don't offer an open API.
Apps coming in iPhone first is by far my most frustrating experience with Android.
The developers of Bounden posted about it here: http://gameovenstudios.com/bounden-on-android-delayed/
Edit: Bounden has a calibration step every time you use the app, and again every time it detects too much drift. If it was a quality issue, they could just warn people that they'll get crummy results if they have crummy sensors.
That doesn't mean you can't do it, it's just easier on the iPhone, which sucks for us Android users.
So while they could likely make and release an Android app, expect only the flagship phones from the last 12-24 months to be supported, nothing a day older (and even a lot of phones within that window still lack the latest API updates).
I've always been disappointed at how/why Google allows fragmentation.
It certainly seems to me that they have the power to force a reunification in Android (and that if they did, the world would cheer).
Why not tell Samsung that Google must determine the update schedule, not Samsung? And if Samsung adds non-standard hardware, they must provide working updated code by this schedule.
And if Samsung doesn't like it, they are welcome to move everything over to their Tizen OS.
Even if Google cannot fix the past, why not do this going forward?
Edit: this is a lot of downvotes for no comments. What's wrong with my comment?
An example of such a condition is that if a manufacturer makes any phone that includes Google apps, they must not manufacture a phone that runs without it. This is why Amazon had such a hard time finding a manufacturer for their Fire tablet/phone. As for skinning, Samsung can only do that because Google explicitly allows it.
Lastly, I feel that someone should have pointed this out instead of downvoting you.
It's similar with music apps, were Android has crappy audio APIs and latency issues. Almost all synths, samplers, etc from professional music companies like Korg et co are made for iOS only.
(there's your answer)
Thank you. Oh, thank you.
Not that I won't, I just hate apps requiring signup when not inherently necessary.
Stuff like this should be coming from Apple. For example, it would be awesome if they created prosumer software for filming with a wireless multi-camera setup and edit it all live on a macbook. would love it if upstart young-Turks like web shows could get closer to the production quality of TV talk shows.
Something that makes the right people go wow.
Wait, is iMovie still a thing?
You can get an iOS laptop today, basically an ASUS Transformer:
http://www.logitech.com/en-us/product/ultrathin-keyboard-cov...
Stabilization is not something that costs $15k, at least at this quality. Yes, it's interesting on a technical level (both with Instagram and MS), but there are low-tech solutions that handle this aptly.
So in essence we're talking about a video accelerator (and presumably decelerator)? This doesn't exactly wow me, but then again Instagram was founded on applying filters to photos, so I should recognize the value of hype (and subsequent audience size).
Playback speed is the UI feature people notice, but it wouldn't be interesting if the image weren't stabilized.
Software stabilization obviously also supplants hardware cost, which makes even that # moot. Then there's the acceleration, which can be done by low-level video prod software already.
I'm seriously not trying to just poo-poo this, it's just very underwhelming as a total package.
Otherwise it's just an echo chamber of P.R.
If the original title begins with a number or number + gratuitous adjective, we'd appreciate it if you'd crop it. E.g. translate "10 Ways To Do X" to "How To Do X," and "14 Amazing Ys" to "Ys." Exception: when the number is meaningful, e.g. "The 5 Platonic Solids."
The $15,000 doesn't seem to qualify.
That being said, you could certainly qualify it as linkbait, but why just take out the number? The simile is ruined, and the title just sounds stupid now. If you're going to edit it to make it not linkbait, change it to something like:
Hyperlapse, Instagram's new video stabilization app
Certainly you can moderate how you see fit, just thought I'd offer my 2 cents. I get annoyed when I come back to the same article I've already read with a different title so the inconsistency is frustrating sometimes. Sometimes arguably better titles get changed to match the source, and then when (someone else decides) the source has a bad title we change that too...
We'll change the title to your suggestion. Thanks!
All: the best way to complain about a bad title is to suggest a better one. We see hundreds of articles and titles a day—the idea of getting all of those perfect is cruel. Therefore we're delighted to get help, as long as the principles are clear: an HN title should be accurate and neutral while staying as close to the article's own words as possible.
The main advantage I see for MS Research is that they enable the full dropping of multiple frames to remove large camera motions such as having a head mounted camera and briefly looking from side to side.
In the Instagram examples, the camera stays on a fixed subject and their Hyperlapse algorithm reduces the shake. I suspect that if there were large camera motions, they would be translated into the final product which could detract a lot from the appeal of that video.
I'm not interested in using something that's a part of Facebook's ecosystem or sharing it with my friends on Facebook.
The main advantage I see for Instagram's Hyperlapse is that it's a real app on the app store. It actually exists and isn't just a research project.
https://www.dropbox.com/s/i2w18xh31aixmhv/Video%2026-08-14%2...
1) Video is pretty smooth 2) So many hatchback cars 3) You traveled pretty far without running into a single light.
I drove one way and back so it's not that much road; plus I was in a countryside area that doesn't have many lights.
This statement is undeniably false.
1. https://news.ycombinator.com/item?id=8048413
Edit: Hyperlapse can do this too. Thank you Wang.
What it looks like the Instagram app is doing is selecting and stabilizing individual frames instead of the full 3D scene reconstruction that the MSFT paper described. EG there's no pop in and out of details like the MSFT demos and there's still a small but noticeable judder in some of the videos.
Microsoft:
"Our algorithm first reconstructs the 3D input camera path as well as dense, per-frame proxy geometries. We then optimize a novel camera path for the output video (shown in red) that is smooth and passes near the input cameras while ensuring that the virtual camera looks in directions that can be rendered well from the input."
Instagram:
"... Smartphones didn’t have nearly enough power to replicate video-editing software, but they did have built-in gyroscopes. On a smartphone, instead of using power-hungry algorithms to model the camera’s movement, he could measure it directly. And he could funnel those measurements through a simpler algorithm that could map one frame to the next, giving the illusion that the camera was being held steady"
On a side not it's kind of depressing / scary / crazy that it's now just common place to release these apps for free.
Indie development if not suffocated already on the App Store has little hope competing going forward.
I can't wait until someone copies it on Android.
Non-timelapse but related - I wonder how long until mobile phone manufacturers start using in something like this https://www.youtube.com/watch?v=VphxfCXu984
Post stabilisation is good but it's so much better to get it right in-camera
http://conversations.nokia.com/2013/09/25/how-nokia-lumia-op...
I think the earlier 1020 etc were thick because of huge sensor size.
I have OIS dSLR lenses but none of them are quite as good as that Sony's Optical Steady Shot linked above - it's almost as good as some brushless gimbals I'm using on drones, which are rock steady. Sony claim it's 13 times more effective than their previous stabilisation.
Personally I've been using free video stabilizer Deshaker for 10 years with similar (or better) results. The downside is two-pass rendering. Upside is that it is fairly configurable and can stabilize existing videos. You can get ok results with fairly terrible source video.
Am I missing something here?
From an implementation point of view, this app doesn't use UIActivityViewController, but instead one they've rolled themselves.
They could in fact have added Instagram to the built-in share view controller but decided not to. This also takes away options of Twitter, SMS, email.
Google for "ffmpeg" "image stabilization".
Instagram is certainly capable of putting out apps for both platforms simultaneously. I don't think they wouldn't fail on that without a good reason.
If you find yourself constantly feeling this... that may be a sign you're not in the most suitable walled garden.
It's great the Android offers so much choice in terms of features and price, but the upshot is that you're not buying a device with an android sized slice of market share from the point of view of app support, for many purposes you're buying a device with a that-device sized slice of market share.
In the old PC days the exact spec of a machine didn't usually matter much. However for mobile devices, the plethora of advanced sensors they come with and the complexity of software capabilities that take advantage of them has made fragmentation a much bigger issue, but really only if you're interested in the most advanced features.
For most phone users it doesn't make that much difference, but if you're interested in cutting edge features the quality of hardware and software integration and application support has become critical. Samsung almost has enough critical mass to establish a stable platform for advanced Android applications, but the problem is it just doesn't have the vision and discipline to do so by actually establishing a stable baseline for the features and software in it's phones.
(the simple answer is 'usually no' because it would mean the porting team would have to be working with the main dev team and the launch delayed until the port was done)