About a week ago. I started a thread concerning the heavy amount of blurr and subject vibration in OSMO Action videos when shooting at low frame rate, double shutter speed. After conducting some tests and researching the subject, I think the problem is Rock Steady makes the video worse at low manual frame rates, which is a big problem if trying to shoot cinematic video (24, 30, 48 fps and a shutter speed twice that value). Below is the original footage that made me scratch my head. Keep your eye on various subject. Do you see them vibrating with every step I take despite Rock Steady being enabled?

I then ran a series of tests in my neighborhood. keep your eye on still objects such as the lightpost.
OSMO Action w/ ND16 filter - 24fps - 1,50th - ISOMax100 - Rock Steady ON
OSMO Action w/ ND16 filter - 24fps - 1,50th - ISOMax100 - Rock Steady OFF - EV0.0


OSMO Action w/ ND16 filter - 30fps - 1,60th - ISOMax100 - Rock Steady ON

OSMO Action w/ ND16 filter - 48fps - 1,100th - ISOMax100 - Rock Steady ON
OSMO Action w/ ND16 filter - Auto Settings ISO Max 3200 Rock Steady ON

Are noticing any differences? With rock steady on and in manual shooting, the the higher the frame rate and shutter speed, the vibration becomes less apparent. With rock steady off or in complete auto, there doesn't seem to be much vibration if any at all. I ordered the Hohem iSteady Pro 2. I'll conduct more tests with that device when I get it. I suspect that using a gimbal with RockSteady Off will allow me to properly capture the lower frame rates I want. Just a shame DJI didn't caveat Rock Steady's shortcoming.  

This video seems to confirm my hypothesis.

Nicely done - thanks again for sharing!

I feel like I was just trapped in a version of Groundhog Day! (g)

The later segments seemed smoother.  BUT I still want to know where the lake scene was filmed?

ff22 Posted at 8-25 15:08
I feel like I was just trapped in a version of Groundhog Day! (g)

The later segments seemed smoother.  BUT I still want to know where the lake scene was filmed?

haha. lake agnes, colorado.

Nice videos thanks for sharing

Interesting post.  I recently purchased one of these myself in the past month.  The footage has been ok, not quite as bad as yours in your lake video.  Though I have started experimenting with a Z-Axis stabilizer I got from Scotty Makes Stuff as I can see the up and down motion of me walking too much.  Now I see, we seem to be going full circle and looking back at gimbals again.  I have an original Smooth Q I guess I will dust off and test with.  It seems to really defeat the purpose of the Action, though I am glad you have brought this up and made me think.  Now I am starting to wish I had just stuck with my Hero5 with ND filters and used the gimbal.  

I posted this video before.  I think it shows Rock Steady is pretty good at dampening bumps.  The image ahead is pretty smooth while you can see the brake/shift lever vibrating pretty vigorously.  



Obviously, it is not perfect.

ff22 Posted at 8-25 19:07
I posted this video before.  I think it shows Rock Steady is pretty good at dampening bumps.  The image ahead is pretty smooth while you can see the brake/shift lever vibrating pretty vigorously.  

Agreed. In one of the test videos I disabled Rock Steady, which really made me appreciate how RS dampened my hand movements. But at the price of the camera flailing all over the place, nothing was vibrating.

Awesome - thanks for sharing

Thanks a lot for the post! This confirms my own testing I did after purchasing ND filters. I'm a big fan of the 180-degree rule, but for the Osmo Action, you have to forget it (except when using a gimbal...).
You are absolutely right with your conclusion, the key to good footage is the shutter speed. What bothers me most, is that you actually can't film with 24fps if you need to follow this rule. 24fps with 1000/s is hurting my eyes. Therefore, I almost only film with 60fps+ these days. Makes it hard to mix with 24fps footage in post... damn. Maybe I should get a gimbal. Too bad my Ronin S is too mighty to stabilize the Osmo Action ;)

Thanks a lot for the tests!. Probably whatever causes this, also causes the 50Hz artificial light flicker with RS...

Hi, there. Thanks for sharing this awesome video test that you had created. Rest assured that DJI will continue to do its best to provide the quality products to our customers. Thank you for your continued support!

Agnes was a pretty common girls name here in Denmark, and i also think a few use to be Danes live in Colorado now.

Chris_E Posted at 8-25 22:36
Thanks a lot for the post! This confirms my own testing I did after purchasing ND filters. I'm a big fan of the 180-degree rule, but for the Osmo Action, you have to forget it (except when using a gimbal...).
You are absolutely right with your conclusion, the key to good footage is the shutter speed. What bothers me most, is that you actually can't film with 24fps if you need to follow this rule. 24fps with 1000/s is hurting my eyes. Therefore, I almost only film with 60fps+ these days. Makes it hard to mix with 24fps footage in post... damn. Maybe I should get a gimbal. Too bad my Ronin S is too mighty to stabilize the Osmo Action ;)

"I'm a big fan of the 180-degree rule, but for the Osmo Action, you have to forget it"

This is one of the reasons I sent the camera back. Maybe they'll fix it someday...

Thanks for sharing

Yes, I too have been struggling with this and my Polarpro lenses while filming mountain biking.

It’s suuuper jittery and unbearable with rocksteady ON at 24 FPS with PPro ND4 and a shutter speed of 1/50.

So today I tested with 60 FPS and 1/120th (some believe RS does better at higher frame rates) no good.

Then I turned RS off and wow, no jitters over the small chatter.

I ended turning manual mode off and running in auto with the the ND4 on.

Sad because RS was thought to be the leader in EIS, I guess that means in auto or on pavement only?

This is really bad and I’m shocked no one has reported this as a bug/enhancement.

Why has Rocksteady become Suuperjittery?

I wonder if other cameras such as full frame mirrorless with IBIS have the same issue? i've never heard it reported.

DJI Gamora Posted at 8-28 08:37
Hi, there. Thanks for sharing this awesome video test that you had created. Rest assured that DJI will continue to do its best to provide the quality products to our customers. Thank you for your continued support!

Appreciate the willingness to listen to customers and improve products, however, DJI should probably be more vocal about the shortcomings of Rock Steady. I wasted a lot of footage on awesome trips that I'll never get back while wondering why everything was vibrating.

DogHasTheWheel Posted at 8-29 21:11
Appreciate the willingness to listen to customers and improve products, however, DJI should probably be more vocal about the shortcomings of Rock Steady. I wasted a lot of footage on awesome trips that I'll never get back while wondering why everything was vibrating.

Hi, thank you for bringing this query for us. We do apologize for the inconvenience caused. Can you please send us a sample video that you had taken during that trip for us to check and further help you with your concern. Please keep us posted. Thank you for your kind and understanding. Cheers!

DogHasTheWheel Posted at 8-29 21:08
I wonder if other cameras such as full frame mirrorless with IBIS have the same issue? i've never heard it reported.

The GoPro has the same issue. Electronic stabilization needs light to work properly (that's why all action cams suck stabilizing the footage at night, besides the bad sensor noise). The Osmo Action is... an Action Cam, I don't know why people insists on filming in a 'Cinematic' way... for that you will need either a tripod or a 3-axis gimbal. is has ALWAYS been like that.

First video seems like the lens was dirty.

DJI Gamora Posted at 8-30 14:26
Hi, thank you for bringing this query for us. We do apologize for the inconvenience caused. Can you please send us a sample video that you had taken during that trip for us to check and further help you with your concern. Please keep us posted. Thank you for your kind and understanding. Cheers!

There is a video of the sample footage in the original post. how would you like me to send you the video?

El Diabolico Posted at 9-2 00:25
The GoPro has the same issue. Electronic stabilization needs light to work properly (that's why all action cams suck stabilizing the footage at night, besides the bad sensor noise). The Osmo Action is... an Action Cam, I don't know why people insists on filming in a 'Cinematic' way... for that you will need either a tripod or a 3-axis gimbal. is has ALWAYS been like that.

https://www.youtube.com/watch?v=viHOHRrOeGg

Yeah. I think i'm going to return my Action and get the pocket

DogHasTheWheel Posted at 9-11 22:05
There is a video of the sample footage in the original post. how would you like me to send you the video?

Hi, thank you for the information given. Let me forward this concern to our designated department. Rest assured that we will provide you feedback once we already hear from them. Thank you for your patience and understanding. Cheers!

DogHasTheWheel Posted at 9-11 22:06
Yeah. I think i'm going to return my Action and get the pocket

I have both ;)

I really like the footage on my OSMO Action but (as others have stated) the jitters at lower fps make the camera unusable. It's nearly July 2020 and the issue of Rocksteady not working with filters still persists.

This is a very old topic. Rocksteady works with high frame rates, therefore forget the 180 degree rule & ND filters unless you’re using a 3-axis gimbal...

So then the OA should stop being used?  unusable. or simply not to your liking?

As far as I can see its as  El Diabolico has said, Rocksteady (and Hypersmooth and all EIS) works with high shutter speeds.

The software compares the sequential shots then manipulates the image to artificially create a steady output, it will basically rotate each frame a bit to match the last.

With a high shutter (and good light), the individual frames will be crisp with little blur making it easy for the software to smooth frame to frame, with a slow shutter the individual frames will be more blurred making it hard for the software to work out how to manipulate the frames.

Filters are specifically designed to introduce the cinematic motion blur, kind of makes them and EIS mutually exclusive. Maybe we will see OIS on future cameras but I guess that would make them more fragile.

itchyeyeballs Posted at 7-7 09:01
As far as I can see its as  El Diabolico has said, Rocksteady (and Hypersmooth and all EIS) works with high shutter speeds.

The software compares the sequential shots then manipulates the image to artificially create a steady output, it will basically rotate each frame a bit to match the last.

As far as I know it uses the internal sensors to detect any movement/vibration and moves the usable area to counteract that. That is why you loose so much from the full sensor readout when you turn on Rocksteady; it has to leave a big margin for cropping on the sides. It also makes sense that frames should be captured as fast as possible otherwise they would have blurring within the frame from those movements/vibrations. If you want long shutter speeds you need to stabilize the system mechanically...

CemAygun Posted at 7-8 14:32
As far as I know it uses the internal sensors to detect any movement/vibration and moves the usable area to counteract that. That is why you loose so much from the full sensor readout when you turn on Rocksteady; it has to leave a big margin for cropping on the sides. It also makes sense that frames should be captured as fast as possible otherwise they would have blurring within the frame from those movements/vibrations. If you want long shutter speeds you need to stabilize the system mechanically...

Your right about the cropping but I don't think it uses sensors. I believe its a software algorithm that looks for common points and lines then rotates, warps and crops the image to match the sequence.

You could do the same thing manually (if you had a lot of time) and a lot of post-processing video editors will do it automatically as well.

itchyeyeballs Posted at 7-10 05:30
Your right about the cropping but I don't think it uses sensors. I believe its a software algorithm that looks for common points and lines then rotates, warps and crops the image to match the sequence.

You could do the same thing manually (if you had a lot of time) and a lot of post-processing video editors will do it automatically as well.

I am sure GoPro Hypersmooth do use sensors (it is officially disclosed) so I recon Rocksteady is the same. And it takes forever to analyze a 4K 60 video even on a powerful computer, so it seems a bit unlikely to me that the camera is doing it "manually". Stabilization part of the video takes no time, it is just cropping and rotating. Detecting the offset is the compute-intensive part. Sensors basically circumvent that computational need...

This is no reason to complain about the Action 2 as a product, EIS works this way in general, it needs high shutter speed to get crisp photos which it then uses to examine for feeding the stabilization algorithm. If you have cinematic motion blur, how do you expect any algorithm or AI (as of now) to understand which points it will base the stabilization on?