DJI challenging GoPro?

[MeanRevert]

Also noticed that osmo pocket price was drastically cut. I think DJI realized that given the choice, customers will buy this one. 

[newfoundmass]

These releases always annoy me because all the usual suspects flood YouTube with glowing reviews in unison, allowing themselves to be used as the marketing arm of these companies. 

[sanveer]

4 hours ago, MeanRevert said:

Lots of videos but I think this dude has the best, comparing the Hero 7 and Sony FDR with the Osmo Pocket Action.  It's pretty clear to me that the Hero 7 has the best image quality (sharper, not as contrasty).

 

 

I am not sure how you arrived at that conclusion.

Regardless of the added crop, the DJI Osmo Action seems to have WAY BETTER stabilization than the GoPro, and even the Sony.

The review above isn't complete. Let's wait for part 2, to see the final conclusion.

[MeanRevert]

12 minutes ago, sanveer said:

I am not sure how you arrived at that conclusion.

Regardless of the added crop, the DJI Osmo Action seems to have WAY BETTER stabilization than the GoPro, and even the Sony.

The review above isn't complete. Let's wait for part 2, to see the final conclusion.

In the above video, check out the side by sides at 10:39 and the audio test at 12:20. The GoPro looks better to me, more detail, less contrasty.

[sanveer]

1 hour ago, MeanRevert said:

In the above video, check out the side by sides at 10:39 and the audio test at 12:20. The GoPro looks better to me, more detail, less contrasty.

You're right. The GoPro has a lot more lens flare, and the sharpening (and contrast) on the DJI needs to be dialled down a little, even in the non-HDR mode. Reducing contrast and sharpening could help the video look a lot more cinematic. 

[androidlad]

The HDR mode is achieved using DOL-HDR, a feature seen on numerous Sony sensors including the ones in X-T3 and GFX100.

Basically using 2 ADCs to generate 2 frames of different gain settings almost concurrently. However rolling shutter will double and max frame rate will be halved.

[kye]

Hmm.. Looking forward to seeing @DaveAltizer review the stabilisation in low light.

This is one of the strengths of OIS (and the X3000) and in the iPhonedo Hero 7 review there's a clip of the X3000 absolutely slaughtering the Hero 7 in testing low light stabilisation.

Very interesting camera though.

[kye]

9 hours ago, DBounce said:

So steady, this is truly the future of image stabilization. Forget IBIS, if this tech can be incorporated into cameras of this size, there is no reason it cannot be added to much larger dedicated cameras. Also, in the video he uses De-Warp. I've never heard of this before, but that should be automatically applied in any camera. Honestly, none of the footage really seemed to suffer from warping, so I'm not sure why he chose to use it on the one clip... perhaps just to show he could?

De-warp is the same as Linear mode in GoPro language, which is really de-fishing in normal camera language.  I noticed a bit of fisheye distortion on the Osmo, but it was far less than on the Hero7, so that's a good start.  The GoPro only has Linear in 2.7k mode, because it will be taking a 4k image and stretching it in software, so IQ will be lost, thus no point in giving us a 4K output file.  The Osmo has a much less distorted place to start, so getting a 4K de-fished output file seems more appropriate.  

IIRC the De-Warp couldn't be used at the same time as RockSteady so you'd have to choose between them.  De-Warp is pretty easy to do in post anyway.

3 hours ago, newfoundmass said:

These releases always annoy me because all the usual suspects flood YouTube with glowing reviews in unison, allowing themselves to be used as the marketing arm of these companies. 

Like all camera reviews, which are in danger of being glowing due to direct or indirect incentives to please our camera brand gods, you often have to read between the lines to see where the downsides are.  iPhonedo does a reasonable job of sharing non-glowing facts when he finds them though.  

I think that the real challenge here is that these cameras are being released so often that it's really hard for reviewers to justify all the time and effort to do a thorough review.  iPhonedo does the best job I think, showing side-by-side DR with waveforms, audio tests with his guitar, finding heavily saturated subjects, and generally stress-testing them.  But you're right that no reviewer really gives you a solid list of all the pros and cons - you have to watch a bunch of reviews and kind of compile that information yourself.

In terms of these YT reviewers being the marketing arm of brands, this is an example of a relatively open market - if we weren't interested in the products we would watch and no views would mean no success in marketing, so it's opt-in marketing.  I've opted in because I plan to buy a camera in the next week and the X3000 was winning over Hero7 for me, and now I have to think seriously about if the Osmo Action is right for me.  I suspect not, but it certainly has some excellent features, and none of these are the perfect camera for me so it's about the tradeoffs and prioritising features.

[MeanRevert]

Ok, I lied. This dude has the best review out there. I found the side-by-side, the 2nd video, especially useful.

 

 

[sanveer]

Any information on how good the HDR is? Someone posted that DJI claims a 3-stop.improvement over the regular video. That's a tall claim. 

P.S.: Other videos say its more like 1 atop. Hopefully reviewers tests things a lot more. 

[JurijTurnsek]

14 hours ago, MeanRevert said:

Also noticed that osmo pocket price was drastically cut. I think DJI realized that given the choice, customers will buy this one. 

Where?

Also, I hope GoPro and DJI switch to the 48mpx sensor for even better EIS (you are not getting true 4K with EIS from the cropped 12mpx sensor). Osmo Pocket is still a more versatile film-making tool unless you need to go near or under water.

[Anaconda_]

The lag on the stabilisation is pretty drastic. It'll be quite frustrating to film anything where framing is important.

Skip to 6:30ish for the 'very slight' (around 2 seconds) delay on the panning/movements.

while people are talking about the Pocket though, I recently handed one to my wife, who's not really interested in filming anything, and she didn't want to give it back. She said it was too much fun and got all experimental with it, pushing it through gaps in a fence etc. To me, thats exactly how cameras in this market should make someone feel. You can take a quick photo on your phone, but you can really have fun with a tiny accessory. I predict the lag on the Osmo Action will take all the fun out of kind of filming.

Just my 2c.

[kye]

Another thing that no one is taking about is that EIS can be done in post production, but OIS can't. 

Yes it's great to have stabilised footage coming straight from the camera, but stabilisation in post has the advantage because the camera can't predict the future but plugins or Resolve sure can.

I think I'll end up with the X3000 and stabilisation in post. The HDR would be great to have thought.

I would have thoughts that anyone on this forum would be evaluating a camera with the whole production and post production work flow in mind.

28 minutes ago, JurijTurnsek said:

Where?

Also, I hope GoPro and DJI switch to the 48mpx sensor for even better EIS (you are not getting true 4K with EIS from the cropped 12mpx sensor). Osmo Pocket is still a more versatile film-making tool unless you need to go near or under water.

When they go 8K they will have 6K MEGA-AWESOME-NOT-MOVING mode and you can de-fish and downscale to a nice 4K imagery from there.

[Anaconda_]

3 minutes ago, kye said:

Another thing that no one is taking about is that EIS can be done in post production, but OIS can't. 

I assumed that's why there's the lag when you turn on stabilisation, it's doing the post work in realish time, so can in a way predict the future as you say.

Also, don't rule out that in camera EIS can use things like gyroscopes to smooth out the image, which you can't do in post. Like that accessory that you put on a hotshoe and it records all your movements to metadata and you can stabilise later using that info. Only here, it's builtin. 

Disclaimer: I have no idea if any of these cameras are using gyroscopse to help with the IS, but it would make it more valuable to do in camera over post.

[kye]

@Anaconda_ you raise good points.

The cameras are looking into the future, but as you say it's only a couple of seconds, but Resolve (for example) can see the whole clip before it decides anything.

They do use gyroscopes (gopro bought one of those companies I think?) but that's kind of a cheat because instead of analysing the image they just blindly follow what the gyro tells them to do, but if you have smart enough stabilisation software then it should be able to overcome that. The gyroscope is a hack to reduce CPU required, it's not a fundamentally better method.

[Emanuel]

Osmo Pocket seems a way more interesting concept...

This copycat seems to me much of an empty release.

 

Leave GoPro alone! That's what I feel like to add. BTW, I liked fairly more the GoPro outcome from those comparative videos...

 

In my country, we're used to say there are people who cannot see a poor with a cleaned shirt.

OK, we all know Nick Woodman is not exactly a poor man but I guess you get the meaning... ; )

As photographer and self made entrepreneur he has all my respect and I am not GoPro user nor his client either!

[Video Hummus]

2 hours ago, kye said:

Yes it's great to have stabilised footage coming straight from the camera, but stabilisation in post has the advantage because the camera can't predict the future but plugins or Resolve sure can.

Actually, I believe having EIS done in-camera when the camera has a gyroscope and accelerometers allows those sensors data to inform the electronic stabilization. That’s the only reason HyperSmooth and Rock Steady do such a good job.

I also think it’s somethingOlympus and Panasonic need to offer in their next cameras. Having sensor stabilized and then advanced sensor EIS on top can get you gimbal performance without the bulk.

It is a fundamentally better method because it would be like having to guess the speed of your car without having an speedometer versus actually having a speedometer. Measuring will always be more precise than interpolating.

[JurijTurnsek]

25 minutes ago, Video Hummus said:

Actually, I believe having EIS done in-camera when the camera has a gyroscope and accelerometers allows those sensors data to inform the electronic stabilization. That’s the only reason HyperSmooth and Rock Steady do such a good job.

I also think it’s somethingOlympus and Panasonic need to offer in their next cameras. Having sensor stabilized and then advanced sensor EIS on top can get you gimbal performance without the bulk.

It is a fundamentally better method because it would be like having to guess the speed of your car without having an speedometer versus actually having a speedometer. Measuring will always be more precise than interpolating.

Take it a step further and save the gyro and accel data alongside the unstabilized clip and let computer software analyze it post (beefier CPU, multiple passes and adjustable parameters).

[Anaconda_]

12 minutes ago, JurijTurnsek said:

Take it a step further and save the gyro and accel data alongside the unstabilized clip and let computer software analyze it post (beefier CPU, multiple passes and adjustable parameters).

Like this? Skip to like 3 minutes for a great preview of exactly what it's doing to stabilise. 

2 hours ago, kye said:

it's only a couple of seconds

But imagine trying to film someone while walking next to them. Those couple of seconds will make it very difficult to keep them framed nicely. While it allows the camera to look ahead in time, the operator still works in the here and now.

[kye]

29 minutes ago, Video Hummus said:

It is a fundamentally better method because it would be like having to guess the speed of your car without having an speedometer versus actually having a speedometer. Measuring will always be more precise than interpolating.

I disagree.  In the case of the gyro stabilisation, what you want is for the distinctive pixels in one frame to be as close as possible to their location in the last frame.  It's a frame matching problem.

In image analysis based stabilisation, the computer looks for distinctive pixels in each frame, works out the movement, then works out how to change frame 2 to better match frame 1, knowing the overall direction of movement from frame 1 to frame 1000.  This method is about matching frames by analysing the frames directly.

In gyro stabilisation, the computer wants to match frames to other frames, but does so by taking frame 2 and modifying it according to some completely different source of data.  That data is very useful and does a good job, but it cannot do a perfect job because it is one step removed.  This method is about matching frames by not looking at frames but by looking at something else.

I think people are seeing that the GoPro and Osmo Action are better at stabilising than the software and thinking that's because it's a better method.  I would suggest that it's not a better method but one that has had far more investment in the technology.  The entire action camera industry now hinges on making the footage as stable as possible.  It's a stabilisation arms race of sorts, and without going to 8K it's the only thing that is different between GoPro models.
Conversely, there is no arms race with post=production stabilisation.  The people who specialise in it are specialists, and the people that do serious post-production workflows and invest int them are people that don't record shaky footage in the first place, or if they do then they buy gimbals and IBIS and OIS and tripods etc.  The action camera market has taken gyro stabilisation a long way further than the alternative.  It's an investment thing, not an image stabilisation thing.

Plus, think about how well IBIS works on a 16mm lens - pretty darn well.  Think about how well it works on a 400mm lens - less well.  Stabilisation in post is just fine at 1000mm  

12 minutes ago, JurijTurnsek said:

Take it a step further and save the gyro and accel data alongside the unstabilized clip and let computer software analyze it post (beefier CPU, multiple passes and adjustable parameters).

Interesting idea.  My understanding of OIS is that there are gyros built in, so any camera with IBIS or an OIS connected lens may have the hardware required.  It probably wouldn't be setup to analyse that data and record, it might simply be used to move the sensor around.

3 minutes ago, Anaconda_ said:

But imagine trying to film someone while walking next to them. Those couple of seconds will make it very difficult to keep them framed nicely. While it allows the camera to look ahead in time, the operator still works in the here and now.

I said "it's only 2 seconds" because I meant that 2 seconds isn't much future visibility for the stabilisation to work.

A 2 second delay on the viewfinder would be almost unusable.

You quoted me out of context