Weapon Gyro Data

[Christoffer Glans]

This is more of a feature request, but with the new higher capture rate of the in-cam gyro to the metadata in Weapon, it would be a great feature if there was a possibility for other softwares like Resolve, Premiere etc. to make use of that data as a virtual cam for stabilization while adding this feature in Redcine as well. It would be a game changer for post production to have this as a set of easy sliders.

 

[Daniel Schweinert]

Would be great!

Btw. same idea was said at NAB about the BMD URSA Mini with it's GPS and 9 axis gyroscope (would work in fusion, maya, lightwave,...) but now nobody talks about it - silence. I also think the GPS was ditched in the URSA Mini afterwards.

 

[Wil Wong]

I can see this being a feature that all cameras will have one day. Imagine, no more gimbals or steadicam. It'll be done in the sensor being stabilized in camera (gyroscope or other tech). Gee...I'd have to trade-up my camera again but that would be worth it!

 

[Christoffer Glans]

I can see this being a feature that all cameras will have one day. Imagine, no more gimbals or steadicam. It'll be done in the sensor being stabilized in camera (gyroscope or other tech). Gee...I'd have to trade-up my camera again but that would be worth it!

You can't just do a stabilization without taking in factors like motion blur. It will never replace gimbals and steadicams but it will help making them smoother and when shot with a fast shutter it might work for some shots. But still, I don't understand why this feature isn't looked upon more? Just look at the steadXP, the capabilities are massive, but no one is making anything. How I wish that I had the programming skills to do it myself, imagine the money I would make ^_^

 

[Björn Benckert]

You can't just do a stabilization without taking in factors like motion blur. It will never replace gimbals and steadicams but it will help making them smoother and when shot with a fast shutter it might work for some shots. But still, I don't understand why this feature isn't looked upon more? Just look at the steadXP, the capabilities are massive, but no one is making anything. How I wish that I had the programming skills to do it myself, imagine the money I would make ^_^

that is only a question of frame rates. Think shooting 500fps 360shutter or such then post stabilization could be applied before frame averaging 10 and deal 2 that would give 10 times better result then stabilizing at 25fps.

I think the big use for the gyrod is possibly not to stabilize the shots but more to get a good track for cg.

 

[Blair S. Paulsen]

Would love to see gyro/stabilization capability more fully realized. I realize it's no free lunch and would expect some resolution penalty, but what a terrific tool to have in the box.

Cheers - #19

 

[Christoffer Glans]

that is only a question of frame rates. Think shooting 500fps 360shutter or such then post stabilization could be applied before frame averaging 10 and deal 2 that would give 10 times better result then stabilizing at 25fps.

I think the big use for the gyrod is possibly not to stabilize the shots but more to get a good track for cg.

True, but I've been doing successful tracking and stabilization on shots shot at 90 degree that works well without any notable motion blur affecting the stabilized image. Thing is that most of the time what makes a post-tracked and stabilized shot not working is that the tracking "wobbles" the image. If you get something without "wobbles" it looks like a steadycam/dolly even from handheld. Just look at what SteadXP is doing, that's high shutter shots which works great from what I've seen. Although I haven't seen high end footage yet. However, it can also be used to perfectly smooth out a dolly or steadycam shot, like Finchers does in his films.

And yes, the improvements to CG workflows is also a very underrated. Just imagine a shot where there's movement all over the place and someone needs to mask everything out before tracking the camera and even then it might not land perfectly.

Would love to see gyro/stabilization capability more fully realized. I realize it's no free lunch and would expect some resolution penalty, but what a terrific tool to have in the box.

Cheers - #19

There's no more penalty then when you take a 6K image and reframe it and end up with a 4.5-5K image, it's still gonna resolve down to 4K nicely.

 

[Björn Benckert]

True, but I've been doing successful tracking and stabilization on shots shot at 90 degree that works well without any notable motion blur affecting the stabilized image. Thing is that most of the time what makes a post-tracked and stabilized shot not working is that the tracking "wobbles" the image. If you get something without "wobbles" it looks like a steadycam/dolly even from handheld. Just look at what SteadXP is doing, that's high shutter shots which works great from what I've seen. Although I haven't seen high end footage yet. However, it can also be used to perfectly smooth out a dolly or steadycam shot, like Finchers does in his films.

And yes, the improvements to CG workflows is also a very underrated. Just imagine a shot where there's movement all over the place and someone needs to mask everything out before tracking the camera and even then it might not land perfectly.



There's no more penalty then when you take a 6K image and reframe it and end up with a 4.5-5K image, it's still gonna resolve down to 4K nicely.

Yes off course you can lower the shutter time and get less motion blur, but the trick is to remain at a end result motion blur of 180 deg but divide out the shake motion blur but not the motion blur of the action or the camera movement without the shake.

And high resolution helps. But the stabilisation in it self eats resolution if you want to remove high amplitude shakes as you need to zoom in and stabbing especially roll movement really eat resolution / softening the image. So best off course is to be as stable as possible in camera.

Worked on some of the best gyro stabilised material I ever seen last week, think Jet fighter planes filming in 300knots perfectly stable. I think for such applications the in camera gyros could provide good data to smooth operator movements etc.

 

[Christoffer Glans]

Yes off course you can lower the shutter time and get less motion blur, but the trick is to remain at a end result motion blur of 180 deg but divide out the shake motion blur but not the motion blur of the action or the camera movement without the shake.

And high resolution helps. But the stabilisation in it self eats resolution if you want to remove high amplitude shakes as you need to zoom in and stabbing especially roll movement really eat resolution / softening the image. So best off course is to be as stable as possible in camera.

Worked on some of the best gyro stabilised material I ever seen last week, think Jet fighter planes filming in 300knots perfectly stable. I think for such applications the in camera gyros could provide good data to smooth operator movements etc.

How can you keep motion blur of the action and not get motion blur of the camera? They aren't separated. If you stabilize a shot that's been shot handheld you have to apply post-motionblur in order to smooth out the movement. I'm not saying that you should plan and shoot something hand held and then stabilize, but there are times when you need that camera movement and don't have the time or gear to get it. If you look at Birdman you can see a lot of these things, most notable on the roof where they couldn't go with a faster shutter. When the shot ends up in the cut, it will probably be invisible when watching it as a whole. If you pixel peep the shot frame by frame you will probably see the artifacts, but who does that among the ones who watch what we do?

Stabilization doesn't eat resolution down so much that you can't get a good 4K output as a result. Yes, if you move the camera too much, but the shot to use should be the least shaky one, then you won't drop more than down to 4.5K or 5K and that works for a 4K output.

For narrative work, all the artifacts generated by post stabilization will be close to invisible if done right, even from handheld shots. Birdman is a good example of this and it's not even shot in high res.

 

[Björn Benckert]

How can you keep motion blur of the action and not get motion blur of the camera? They aren't separated. If you stabilize a shot that's been shot handheld you have to apply post-motionblur in order to smooth out the movement. I'm not saying that you should plan and shoot something hand held and then stabilize, but there are times when you need that camera movement and don't have the time or gear to get it. If you look at Birdman you can see a lot of these things, most notable on the roof where they couldn't go with a faster shutter. When the shot ends up in the cut, it will probably be invisible when watching it as a whole. If you pixel peep the shot frame by frame you will probably see the artifacts, but who does that among the ones who watch what we do?

Stabilization doesn't eat resolution down so much that you can't get a good 4K output as a result. Yes, if you move the camera too much, but the shot to use should be the least shaky one, then you won't drop more than down to 4.5K or 5K and that works for a 4K output.

For narrative work, all the artifacts generated by post stabilization will be close to invisible if done right, even from handheld shots. Birdman is a good example of this and it's not even shot in high res.

Yes you can separate the two I explained a few times above how to do it. The more sub frames you can shoot the the better the result.

So one more time:

shoot 100fps... 360 shutter.
In post stabilise.
After stabilisation in post do frame average over 2. (this gives you 50fps 360 shutter where the camera move was stabilised prior to getting half the motion blur)
Now take the result, the 50fps blended from 100fps 360 cut, and speed 200% no interpolation, or simply extract every second frame.

This gives you a 25pfs 180 deg clip with half the rolling shutter and camera shake motion blur problems as if you shot

The result of the above will give you twice as good result then shooting 25fps in terms of having a correct 180 deg shutter at 25fps.

Now apply the above to phantom 4k and shoot 1000fps.... that gives you 10 times better result than shooting 25p 180 deg shutter. If you remove the shake before the fame average and deal 2 is applied.

So by stabilising, and then after stabilising doing frame average of 360 sub frames does indeed handle the camera shake motion blur and action motion blur separate. and with the right formula you get right down to 180 deg shutter.

Also if you use phantom or any such fast camera that can give you a huge load of subframes, you can be even more elaborate so with simple scripting you can set your motion blur to be more of the sine wave kind instead of ugly square curved (on,off) shutter. Simply by making a script that weights the subframes differently.

For sure I see a not to distant future where frame rate and motion blur will all be is set in post. This will come when the sensor are light sensitive enough to shoot at pretty much any shutter speed and the data rates are less limited. Shoot 3000 fps or such and then frame average to get the motion blur and frame rate wanted. Such process also allow for super resolution from multi sampling and is also a treat for sorting sensor noise etc. But we are not at 3000fps at the moment, but the metology works at 100fps just the same only that you just get a sample of two.

On birdaman they shot alexa if I´m not mistaken. I think alexa do not do 360 shutter at high frame rates. So that would give you a glitch in the motion blur when doing the average. even so I think they would avoid this tactics as they where on steady pretty much all through and then they look more to stabilise low amplitude long frequency. Where as with hand held you short frequency and possibly high amplitude shakes... thats what affects the motion blur.

I´m sure Graeme is allready quite far down this rabbit hole. The fact that the camera has both frame summing and frame average and gyros does not really make it such a big leap to get to very advanced in camera raw processing. Like sub frame recording, stabilisation and frame blending before the footage even land on the mag even though I think a post process is possibly a better solution.