The state of editing software for Gear 360

There are two main programs for editing 360 video: Video Stitch and Kolor Autopanovideo.  Both are about $750 for a license.  Both offer a trial demo version that puts a horrible watermark all over your video.

Video Stitch is an interesting program because it offers Nvidia acceleration.  It "wants" two single spherical videos, instead of what the Gear 360 offers, which is a side by side dual spherical video.  So there would have to be an extra step in post production to split the dual spheres into two separate files, then stitch those, and there was a guy who showed how to do this on a Youtube video, but I lost the link and lost interest in finding it again, since it seemed like a lot of extra work in Adobe Premiere.  I experimented with writing a script in AVISynth to do the same thing, but lost interest after I could only open the audio of an mp4.  In addition, my version of Video Stitch barfed when fed Gear 360 video and I tried downloading the latest version, etc.

I think these programs were designed for real estate photographers wanting to put 360 panos on their listings.

Adobe Premiere Pro has a 360 viewer but no editing tools.  They have this irritating licensing scheme now that costs $20 a month, but this means you can try out the software for a month with only a $20 investment, better than spending $600 just to find it doesn't do what you want it to do.

I bought a MAGIX license on the promise of 360 tools but there are none, at least in the version I downloaded.

Samsung was kind enough to upgrade ActionDirector to stitch 360 video from the Gear device.  Here's a link to download the latest version.

The stitching is extremely slow even on an i7 with no video card.  I added a GTX 970 card and got stitching that was slow but not ridiculous.

On ActionDirector, YOU CANNOT FIX HORIZON ISSUES IN POST, at least as far as I can see.  The stitching software is set up to assume the camera is perfectly level at all times.  You cannot adjust for the horizon that I can tell.

Video Stitch offers horizon fix but the demo craps out after loading Gear 360 video at least on my rig, with the latest version.

ActionDirector for Gear 360 isn't "bad" because it offers the three basic functions you need to work with 360 video: stitch, render, and upload, all within one program.  Postproduction is therefore pretty simple.  Just be sure the camera is level!

Actually the workflow is pretty simple, drag raw mp4s and then wait for stitching.  Come back and cut video, then produce.

AD offers automatic upload to Youtube within the program, but you don't really want to do this because it doesn't leave a local finished video product on your PC, so if any step of the process dies, you have to start all over again.

The final step before you can upload rendered video to Youtube is to "inject" metadata telling Youtube it's a 360 video.  Just giving it a cool filename like "my 360 vid" isn't enough!

Everyone seems to use a program called Spatial Metadata Injector (creative!).  I've linked it here.  It's pretty self explanatory.

So here's the official workflow:
Shoot video->transfer video to PC->Open 360 ActionDirector->Import video files into 360 AD program->wait for stitching->drag imported clips to final video timeline->cut out bad video clips->"Produce" final mp4 video->Inject metadata->Upload

It sounds like a lot, but it's actually not that bad.  The steps all take a while because the files are really big, but otherwise there's not a lot of human intervention needed.

On being an "early adopter" pertaining to 360 vid

I have been an early adopter in the past.  I experimented with Linux in the early 2000s.  I was fascinated with wifi when it came out.  I had an early Palm Pilot and endless derivative gadgets.  I tried the Psion 5 and Phillips Nino.  I still have a Kodak ZI8 HD video camera--and a cell phone HD camera that does basically the same thing and a bunch of other stuff.
It's a pain in the neck being an early adopter because much of the time you end up with an overpriced paperweight.
Still it's fun to get to use the technology early.
The main use of 360 video, to me, is as a tool to preserve vacation memories.  Sure a phone camera is fine and probably your phone will have some sort of 360 camera on it in the future.  But today, with every trip you take, you can preserve some immersive video.  That to me is more valuable now than waiting until later.
I think 360 photo video is a pretty big step forward compared to conventional photos, in the history of photography.  I have heard the Gear 360 compared to a handheld drone, which I think is not a bad way to think about it.
I haven't gone the drone route because I think they're invasive and distracting.  My whole goal with 360 video is to relive a memory immersively.  It's hard to do that if your gear is attracting a lot of attention because of noise.
Google Cardboard is such an awesome invention, and I really appreciate Youtube's support of the format.
I think compelling travel 360 videos will have an impact on travel.  I worry people would be more likely to stay home, but perhaps they would be more stimulated to travel.  The next generation of Rick Steves and Anthony Bourdains will find a way to incorporate this new technology.
Human memory itself is so imperfect compared to even the video and viewers we have today.
There are not a lot of videos on "Hey guys, here's how to shoot 360 video" because I think there aren't too many people who know how to do it well.  It's like a new musical instrument.  Guitar has been around a long time, so there's plenty of people who do it well, and who can teach it.  360 video has only been around a few years at most, and most of its practitioners are real estate video people.  Clearly there's going to be a long learning curve.  Social acceptability is also still around the corner.  Giving out Google Cardboard out with the New York Times was a nice stunt, but it hasn't sparked a huge mindshare so far.
YouTube will take video up to 128Gb.  Facebook says they will only allow up to 1.75Gb, but they may take more.  The video upload may be the most painful part of the whole process because it's pretty slow.
TimeWarner in LA recently upgraded internet and now offer a home product with 20Mbps upload speed.  An hour of 360 video is 30Gb, so the faster upload speed makes a big difference.  Never did upload speed matter so much.
It's kind of like the olden days when you had these multi-Mb downloads on dialup, except now you don't have to tie up the house phone line for hours sending or receiving files.
Facebook allows more granular control over who sees the videos.  Youtube gives all-or-none permission.  Youtube offers the goggle experience.

Hardware

This blog is evolving thoughts on 360 video, what it means, implementation
This first post will talk about hardware so far.

360 video is a journey into the forefront of technology.  It's doable, today, with tools new and old, but you will need to know a few things to have a successful journey.

360 video is in early adopter phase now.  It will probably take a long time, if ever, to become mainstream because of the hardware and patience it requires.

I've committed to Gear 360 at this point.  People could probably argue for or against.  The few voices out there who care about this sort of thing seem to agree G360 is the most advanced small consumer camera out yet.

1. Cameras
Gear 360 is clearly the best camera out on the market so far.  Resolution is acceptable, form factor is small, "it just works."  Price is not astronomical.  Example: https://www.youtube.com/watch?v=hocIatxuddk  Samsung seems committed to VR.  You can edit video on your phone, not that I'd ever want to do that, but perhaps you might.

Your alternatives:
* Kodax Pixpro 4k - have to buy two cameras. Bulky. Software looks pretty intuitive.  You need to stitch on the PC using two separate files.  Expensive.  Resolution is pretty good.  Fry's might have one in stock.

* 360fly - not really 360, big blind spot on the bottom.  There are a few of these type of products.  It's not really immersive to see a big disc logo at the bottom of the image.  You can buy this in Best Buy.

* Ricoh Theta S - lower resolution.  Really unobtrusive form factor.  In-device stitching (!)  Can't expand the memory.  I don't think this camera is one to get until they upgrade it to 4k and expandable memory.

* An array of GoPro's - super expensive, very complex postprocessing.  Excellent resolution.  Example: https://www.youtube.com/watch?v=zDAYZU4A3w0

* Nikon announced a 360 camera at CES but so far no hardware.

* The DJI Osmo is also a really cool piece of kit, but the lens isn't all that high field of view, battery life is short, and it's more expensive than the Gear 360.  It's not really not an immersive VR camera.

* How about a GoPro Black 4k or Sony 4k action cam?  Picture quality is sweet, but will Youtube accept a 121 degree field of view video for Cardboard upload?  I doubt it.  There are tools for viewing "conventional" action cam vids on a set of goggles, but it's not really 360.  So far neither of these companies are committed to 360 video.  Perhaps we'll get a 360 Sony camera with the Playstation VR hardware?  This is the difference between Sony and Samsung--Samsung seems more committed to immersive video with the Gear 360 and Gear VR combo, while Sony is taking a gaming tack with VR.

Realize the resolution will not differ a whole lot between products at a given price point because the underlying hardware is probably pretty similar.

2. Video card

You will need a graphics card to stitch video.  A "real" graphics card.  I had a $49 graphics card already installed with an AMD GPU, which doesn't help with stitching.  There's probably a reason why stitching requires hardware acceleration with an Nvidia processor, but don't ask me what it is.

I went out to Fry's ended up with a GeForce GTX 970 with 4Gb RAM.  I already had a pretty high-end desktop Lenovo i7 with 12Gb RAM, but it was stitching very slowly.  I use the base Gear 360 software you download to stitch the video.

The Gear 360 gives you a video consisting of two globes, one from the "front" camera and one from the "rear" camera.

I'm not sure what the software is doing so slowly--the stitching isn't excellent.  Color matching is not great and close objects definitely disappear in the stitch line.  Also, it's not like the stitch line can move a lot during the clip, so frame by frame the stitching doesn't vary much.  The system should analyze a few frames of the clip, decide how to stitch the two images together, and then do it.

As an aside, the Gear 360 stitching software, ActionDirector but in a special version for Gear 360, has a serious memory leak problem.  Leave it running overnight and don't expect control back over your computer until late the next day.  

By the way, the 970 is not at all the highest end graphics card out there, but it has a reasonable number of CUDA cores for the price.  Note, I'm not exactly sure what CUDA means, but it sounds like the more cores the better.  Has anyone benchmarked stitching 360 video versus graphics card prices?

Of course Nvidia just released the 1060 but it's hard to find, a little more money, and not clearly better performance than the 970.

3. Power supply

The GTX 970 wanted a 500W power supply, so I swapped out the OEM 210W PS that came with the Lenovo.  I'm not sure I want to find out what happens if you run the wrong spec PS on a higher end graphics card.

4. Phone linkup/viewfinder

It can be a little weird thinking about a camera with no viewfinder, but I think a 360 cam is distracting enough without having to worry about a second device to see the view.  Therefore I use the camera with no viewfinder.  I got the Gear 360 software working on my trusty Galaxy S5, but it doesn't work too well.  I suppose I might as well upgrade to an S6 or S7 but they're still pretty expensive and I just can't bring myself to drop $650 on a phone that might easily end up in a toilet or in the ocean.

Really, if the camera is seeing everything around you, why on earth would you need a viewfinder?  You can't even see half of what it's shooting anyway with any conventional viewfinder, so what's the point?  The only two variables with this camera are keeping the horizon level and figuring out how high you want to hold it, neither of which require a viewfinder.  You may disagree but that's my thought.

5. Hard drive

This video eats up a lot of HD space.  Figure 3 minutes per Gb of raw, unstitched video, and another 3 minutes per Gb stitched.  Then you have to produce the stitched clips into a final .mp4.  Figure on 3Tb SATA for enough scratch space for editing.