DJI Osmo Action VS Osmo Pocket - RockSteady EIS VS 3 Axis Gimbal - Stabilisation Test And Comparison 4K 25 FPS
Here's a video showing how the Action and Pocket compare with each other with regard stabilisation. Basically this is an Osmo Action vs Osmo Pocket stabilisation test and comparison. Where I'm comparing and testing the RockSteady electronic image stabilisation EIS of the DJI Osmo Action against the 3 axis gimbal of the DJI Osmo Pocket.
As each stabiliser or stabilisation method is uniquely different, this video should give us the opportunity to see what those difference are. Between the two stabilisation methods it's maybe not really a case of which one is best but rather which one is best suited to a particular shot, movement or shooting scenario.
Both cameras are set to be as similar as possible with the obvious differences of aperture, focal length and FOV. Both are fully auto for exposure, auto white balance, D-CineLike gamma, the Osmo Action has Dewarp switched on and both set to 100Mb/s. I changed the gimbal and focus modes of the Pocket in certain shots to better suit the shot.
This comparison will also be very good for gauging other differences like sharpness, colour, exposure, internal encoder quality and a bunch of other stuff.
For sharpness, compare the centre of the Action's picture as this is less effected by the pixel manipulation that occurs with its picture processing that will inevitably cause blurring. This is not a down side or issue with the Action, it's just simply how it and similar cameras work (GoPro 7 etc.) with its EIS and geometry correction.
On the point of sharpness. I'll be doing a video soon where I use only the 4K UHD section of the Osmo Action's sensor for gauging native resolution, basically crop into its full sensor resolution of 4000x3000 and only use the centre 3840x2160 pixels. This will result in a slightly barrelled picture due to the focal length of the lens but will be a 1:1 pixel readout of the sensor, assuming other processing and picture/pixel manipulation is bypassed.
In this test both cameras where recorded and edited in 25 FPS in 4K/UHD. For the YouTube upload I used the Grass Valley HQX production codec set to maximum, which is used within my NLE which is Edius, also by Grass Valley. This is a 10bit 4:2:2 YUV codec set to rec.709 in this particular instance. What's important to note here is that I've done many production and quality tests with both the Grass Valley HQ and HQX codecs over the years for feature film and TV production and they are in my opinion the best production/intermediate codecs of their types.
The reason for stressing this point is because the master HQX file is visually indistinguishable form the source files of the Action and Pocket. Which means that the YouTube encode is as visually good as it would be if using directly uploaded native files from both cameras, which would be impossible for such a test as there are overlays and not just straight cuts, which means any type of smart render couldn't be used.
Basically, this is the best possible quality you can get without uploading an uncompressed file, although it's worth noting that HQX is visually indistinguishable compared to an uncompressed YUV version anyway. So and despite the heavy YouTube compression, this is probably about as good as it gets for comparing such things. BTW. The upload file was about 32GB.
Last thing, honestly Nothing was touched in post, what you see here is exactly how both cameras performed in what was essentially fully auto modes for both. Aside from a dodgy operator and one massively badly positioned shot, for which there was no cover for without editing the music. The shots from both cameras represent what anyone can do with either the DJI Osmo Action or Osmo Pocket 'straight out of the box'.
Subscribe or call back for more DJI Osmo Action and Osmo Pocket videos or visit my DJI Osmo Action video website: http://djiosmoaction.com/
Hello and good day Dave. Thank you for sharing this nicely done comparison video you have created. You did an amazing work and thank you for your valued support.
Hi David.
During the side by side comparisons, I assume you did two walkthroughs rather than mounting them alongside each other. I say that because the light levels seem to differ between the two.
Also on one side by side clip, the Osmo Pocket side is shooting mainly shadow, making it seem slightly softer than the Osmo Action side.
Apart from that and taking the differences aside, I was impressed with the colour being roughly the same making matching skies between the two slightly easier if switching between the two for a spliced shot.
You are convincing me that I need a set of ND filters at some point as shutter speeds seem quite fast in both examples.
I found this video quite useful.
Thanks David.
Ray-CubeAce Posted at 5-24 13:21
Hi David.
During the side by side comparisons, I assume you did two walkthroughs rather than mounting them alongside each other. I say that because the light levels seem to differ between the two.
Also on one side by side clip, the Osmo Pocket side is shooting mainly shadow, making it seem slightly softer than the Osmo Action side.
Hi Ray.
Both mounted on a grip system side by side. While these obviously aren't gen-locked, I used a clapper to sync the starts so they are maximum half a frame out of sync.
All the movement and positioning differences you are noticing are the differences between the two stabilisers. Rocksteady does a fair bit of moving and as both may be compensating in different directions at any one time, the difference in position between them becomes more exaggerated. Plus the Pocket does go off real quick and on long takes will drift and need centring. The shot near the end with boats/barges in the dock was my fault, I didn't centre the Pocket so it was off all over the place compared to the Action.
Colour and exposure were 'as is'. I didn't use any exposure compensation or touch them in post as I wanted to give an exact 'out of the box' in auto exposure for both. The other thing that would tip the scales between them as far as colour is concerned will be their auto white balance, difference in their fixed aperture and their lenses and possibly their sensors.
Because it was very bright and in auto, they will have, or should have selected their base ISO 100 and closed down with a very high shutter, which as you've said, will alter the temporal characteristics of movement. I believe they are also native ISO 100, although I can't say for sure.
I've got some very interesting tests and results in other areas that I'll post at some point.
David, you could have shown this in 1 minute, 5:20. I don't understand why people make these things with long bouts of music with endless images that look so similar, you can't tell anyway.
DJI Stephen Posted at 5-24 07:45
Hello and good day Dave. Thank you for sharing this nicely done comparison video you have created. You did an amazing work and thank you for your valued support.
Hi Stephen.
You are welcome.
I'll be doing more stuff with the Action. I can't remember the last time I was so impressed with a camera, I'm having so much fun with the Action.
I take it the HDR function is all in camera processing? However it's done and regardless of if it's true HDR or not, I think this is the killer application for the Action even though Rocksteady is awesome.
TheCameraGuy Posted at 5-24 14:14
David, you could have shown this in 1 minute, 5:20. I don't understand why people make these things with long bouts of music with endless images that look so similar, you can't tell anyway.
Because for those who know what they're doing and know what they are looking for, longer and varied examples give far more information.
You obviously don't understand such things as you say the images were so similar and that you can't tell anyway, when the images obviously weren't so similar and you could tell. For those who know what they are doing and what they are looking for, there was a lot of different information across those clips.
So maybe go and do the type of video that you think would appeal to people like yourself who don't really know what they're doing.
You are very much welcome Dave. It is great to know that you are having fun with the DJI Osmo Action. We will certainly wait and watch your review and test videos. Thank you for your valued support.
Both mounted on a grip system side by side. While these obviously aren't gen-locked, I used a clapper to sync the starts so they are maximum half a frame out of sync.
Hi David.
Thanks for the in-depth reply which makes sense to me.
One other question as you mention the problem of gimbal shift with the Pocket.
Was this using FPV or tilt lock mode?
As for being too long?
I would have liked an extra section with the Osmo Pocket and Action showing a previously shown clip but with the camera sides shown reversed so the Pocket was to the left and the action to the right. That way I could have a direct comparison as to the differences in how both performed in the shadow areas in the same clip.
Last point.
You are right. It is the temporal characteristics of the pan I was watching in the dockside clip but noticed the smooth movement of the vehicles traveling in the same direction in the background. Hence my thought of the use of ND filters which I've never used in the past but can now see the benefits of using.
Excellent work David and looking forward to seeing more tests.
Thanks for the in-depth reply which makes sense to me.
Hi Ray.
I tend to have more drift issues in follow but the other two do it as well. This is my second Pocket and I didn't notice it as much with my first, maybe this one is a little more susceptible.
I'm going to do some locked off comparisons at some point. That may be better for gauging certain stuff.
Totally agree with using NDs to shoot 180 degrees. The only problem I would have had in this test was the constant exposure/light changes etc. The things that TheFunnyGuy didn't understand or notice. Because of that I would have been stuck to all single angles after exposure was set.
I appreciate you can go auto with less varying and slow shutter with a fixed ND when changing position and light in one take. But unless you're shooting 180, you'll still get issues when panning or countering the direction of the action.
Have you ever seen the constant variable Sony ND on some of their cameras? It's like magic. You can set a 180 degree shutter to any frame rate and then tell the ND to constantly compensate the light, it's genius. You can track in and out of light and have constant exposure and iris and constant DOF.
I think they introduced it with the one of the FS5s and I think it's on the FS7 MK2, maybe others. I'm a massive Sony fan, wish I could afford one of those cameras.
Check out this video if you've not seen this type of ND before and do some YouTube searches for examples.
I tend to have more drift issues in follow but the other two do it as well. This is my second Pocket and I didn't notice it as much with my first, maybe this one is a little more susceptible.
The concept of a constant ND is new to me David. 2 to 7 stops is very impressive.
So is it a motorised affair or purely electronic?
I think any counter pan to movement is difficult which is where focus tracking comes in handy. It seems to do a better job than manual panning. One good reason for using a phone with the Pocket as opposed to many against such use I guess, as quickly drawing a box around something you want to track is much easier and quicker with the phone screen than trying to use the Pockets screen.
This is a really informative video, thanks for posting it. What surprises me is how good RockSteady is compared to the gimbal. I have both cameras but haven't done a direct comparison like this. I suppose the Pocket would be able to compnsate for a bigger range of movement than the Action, but this demonstrates that they're very close to equal when compensating for walking movement.
BigStuart Posted at 5-25 02:15
This is a really informative video, thanks for posting it. What surprises me is how good RockSteady is compared to the gimbal. I have both cameras but haven't done a direct comparison like this. I suppose the Pocket would be able to compnsate for a bigger range of movement than the Action, but this demonstrates that they're very close to equal when compensating for walking movement.
Hi Stuart.
Yes, I'm also quite surprised at how well Rocksteady performs.
If I had to say what the differences are, it'd be that the gimbal is better for smooth slower movements and RS is better for faster jerky ones.
Although they're both great on the slower more controlled movements, I feel that the gimbal does work a bit better. On faster more random jerky movements, typical action stuff for instances, I'd say that RS has the upper hand as the motors/torque of the gimbal can't respond as quickly and won't be able to go beyond each axis' end stops.
I'd also say that anyone who thought that DJI's introduction of the Action would interfere or diminish the role of the Pocket, can breath easy. These are definitely two different cameras and while they do have a certain amount of cross over, they are video cameras after all, they both better each other in different areas of shooting and each is more suited to certain shooting styles compared to the other.
lvl.4
DJI team
lvl.3
Nothing was touched in post, what you see here is exactly how both cameras performed in what was essentially fully auto modes for both. Aside from a dodgy operator and one massively badly positioned shot, for which there was no cover for without editing the music. The shots from both cameras represent what anyone can do with either the DJI Osmo Action or Osmo Pocket 'straight out of the box'.
Captain
.