Hero4 stereo rig with CV1

Check out http://freedom360.us/camera-settings/
It provides info.
Highest quality is 2.7k 4:3 (8k total res)

The GoPro Blacks provide the widest range of options.

>Which Hero4 resolution/fps would better match the CV1 resolution?
1440 or 2.7k should be fine but 2.7k gives you higher res.
fps - You're trading off resolution vs fps but 30fps is fine for live action. (60fps gives more live video sense) The headset refresh rate needs to be high to reduce lag but the fps of a video doesn't impact that.

>post-sync the 2 cams in software
Keep in mind even if they we're synced you still have to post-sync so they start at the same time.
But post-sync only gets you +- 1/2 frame since the cameras are running wild and no post-sync will fix that.
Some people shoot at higher frame rates 90, 100 or more simply to get closer sync and then do a fps down conversion. As noted the trade-off is lower resolution for faster frame rates.

"scottsquires" wrote:

Some people shoot at higher frame rates 90, 100 or more simply to get closer sync and then do a fps down conversion. As noted the trade-off is lower resolution for faster frame rates.

Thanks for the feedback. That workaround is very interesting. Let's say i initially shoot at 960p/100fps to achieve closer sync. Which software does the fps down-conversion then? Will resolution get better with this fps-down conversion? I am trying to understand the reason for doing it. Btw i use Kolor Autopano.

According to Freedom360: "Do not use SuperView Mode or the 1080P mode. It is not stitchable. Aspect ratio of the individual videos needs to be 4:3 in order for the six videos to have enough overlap to be stitched"
Do these rules apply only for 360 videos or for stereo video (two cams) as well?

FPS change -
As j1vvy mentioned in this thread https://forums.oculus.com/viewtopic.php?f=33&t=24748
you could use every other frame.

Most editing software can do some frame rate adjustment. After Effects can do some and certainly Nuke can do much more refined fps changing. (using optical flow and other methods)

>Will resolution get better with this fps-down conversion?
Not really. It can reduces noise in the image if it's merging the images but if it's only 960 pixels you're unlikely to get much additional info although you could do a resize at the same time depending on software.

> I am trying to understand the reason for doing it.
If you're moving the camera rig or have a person walking through a seam line or throw a ball across a seam line -
If the cameras had true sync all of those seams would match temporally. But since they aren't synced you can only get it to within an integer frame of each other. So that's possibly +- 1/2 frame. If you don't have things cross the seams or they are are very slow it may not matter.

Assuming 180 shutter at 30fps that's 1/60th second. (GoPros can't set shutter since they're autoexposure so that will vary with your lighting) So 1/2 frame would be off 1/120th second. 90fps gives you finer temporal resolution.
90fps shutter is 1/180th second so 1/2 frame would now be off 1/360th second.

As with all of photography everything is a trade-off. You have to determine whats' the best settings and workflow for your specific project. Is it fast action cross the seams where it's going to make a huge difference or do you want to squeeze out more resolution?

>Aspect ratio of the individual videos needs to be 4:3 in order for the six videos to have enough overlap to be stitched"
Do these rules apply only for 360 videos or for stereo video (two cams) as well?

This only applies to their 360 rig. All the multiple camera rigs require overlap (typically 2% or more) of the images so the software can line them up and merge them together to look like one image. 4:3 makes it more efficient with less cameras. If you want to do a different aspect such as 16:9 then you will need more cameras. Think of having to cover a given area and having small pieces of paper cut in 4:3 or in 16:9 shapes. 4:30 will cover with less pieces.

Stereo only requires the views to overlap to simulate what the eye sees. No need for stitching. But of course 16:9 will give you less vertical view than a 4:3 image.

That 2% is meant to be 20%. The more overlap the better typically because the software looks for matching points on each image.

Keep in mind the final output will be less than you record in almost all cases. The Gear VR and other systems have limits as to resolution and fps they can playback at. So match the settings to the recommended hmd setup.

Also note that when you upload to one of the many distribution sites (such as YouTube, Vrideo, Little Star, etc) they will likely recompress and adjust the video to allow streaming, etc. (i.e. you'll lose resolution and settings will likely be changed by the service)

Best is to do some simple tests for your particular needs before committing to a major project.

@scottsquires Lots of interesting stuff, thank you for taking the time to explain in detail.

The only certain thing newbie or not, is that trying to sync without a cable (aka genlocking) can cause huge frustration and can easily lead to broken 3D. And that is why i am re-considering the dual hero system..:)

Another important thing i am concerned about is the interaxial distance between the two lenses. And ofcourse that varies from project to project and items distance.. BUT what would you guys choose if you could only have one interaxial to produce immersive 3D for the grand majority of the IPDs outhere (children excluded)?

Typical IPD for adults is from 54-68mm, what would be your sweet spot for covering regular events, vacations etc?

If you're focus is on 2 camera stereo then look at the Hero 3 or 3+(?) if you can get them since GoPro actually made a stereo rig with a sync cable for 2 cameras. Take a look at http://lucidcam.com It's a 180 stereo camera system but won't ship until end of the year. Yo can also do a search for stereo gopros since there's quite a few forums and articles about this specifically.

Interaxial - Different interaxial while filming will change the amount of stereo as it were for aesthetic purposes and is used frequently to compensate (such as making it wider for telephoto). If you're aiming for headset stereo then you want it to be average of close to average interaxial so it looks real. It will likely still look real to those with different IPD but with slightly exaggerate or subtler stereo cues. Many simple rigs or stereo cameras, such as the GoPros, are locked. 65-67mm IPD is probably the target for most. 95% of men fall in at 68 and 95% women are 64 if this is correct https://en.wikipedia.org/wiki/Pupillary_distance

In the end the question is what you want to do. Stereo with 2 cameras can carry close to 180 degrees depending on the lenses. This works fine for center and depth will fall off toward edges. You can do 360 stereo but currently many compromises and difficulties getting good stereo. Or you can do 360 mono for a full surround with a chase of doing high quality images.

Standard interaxials will not look "natural" if things are closer than 1m. They wont look natural because you wont be able to converge them comfortably. When I am shooting close I generally use an interaxial of 50mm. Similarly current HMD resolutions are so low that distant subjects will have no discernable depth with standard interaxials so you have the choice of a flat impression if you stick with standard interaxials or scale distortions with a larger than standard interaxials with a mostly distant scene. I often use an interaxial of 90mm for mostly distant scenes. With the Dual Hero System for Gopro 3+ Black you can get interaxials from 30mm to 90mm by cutting away at the plastic of the Dual Hero System backs and rerouting the linking cable path. As scottsquires suggests with 180 fisheye pairs depth falls off to the sides -- but you can use this to configure the camera position -- so that very close scene components can be positioned to the sides, for example.

Are there any Firmware hacks for the GH4 that can capture a square video?

I want to use circular fisheye and not waste bandwith saving the black portion of the frame a d hopfully get a higher frame rate.