DJI Mini 3 Pro video stutter?

[MaximusCZ]

I love learn, people here too. I got history with trying to match monitor with gpu output frames for some time and one does not survive intact

 

[Remote Pilots Australia]

Sorry, you're right. Frustration and exhaustion explaining why YT streams can not be used to judge pretty much anything about the quality of a video, for at least a dozen technical reasons. @MaximusCZ covered a good few of them. He clearly has not gone around and around on this one for years.

I was gonna keep my mouth shut, and should have

Good job to acknowledge human frailty and graciously done. No respect lost here.

 

[mroek]

Good job to acknowledge human frailty and graciously done. No respect lost here.

Absolutely, I agree. I wasn't aware of the history here.

Technically, all videos are streams, be it from YT or your own SSD, but the latter will of course be able to deliver data much faster. However, if no frames are lost between YT and your browser, data loss should not be a cause for video stutter or lack of smoothness. Differences in video frame rate and screen refresh is also a problem that should be independent of the source of the video stream. The same is true for shutter speed and motion blur, where a high shutter speed (compared to the frame rate) can cause video to look less smooth due to lack of motion blur.

I am just trying to understand the underlying mechanisms here, I'm not trying to step on any toes. Video encoding and playback is of course a wide and complex field, but as a computer programmer I do believe I am able to at least understand arguments that are well reasoned (as presented by @MaximusCZ in an earlier post in this thread).

I have no problems agreeing that properly judging video quality from any drone/camera requires access to the original footage. However, if the quality is good, and the author hasn't done errors or dumb choices in post, it is entirly possible to make the video look very good on YT as well.

 

[leebroath]

That's why I facepalm every time there's some debate whether to go 24, 25 or 30 fps, and everyone is talking about motion blur only. Heck, its clear as day. In the Lighthouse video you can see the guy is recording at 4k 30 fps, but if you right click the video and check "stats for nerds", you see YouTube is actually serving 25 fps video. PLUS its playing on your 60 Hertz screen, so 60 fps. You record 30 fps, throw 5 frames away each second to get 25 fps and then play it back on 60 "fps" monitor. Of course you are gonna see stuttering, you have 25 frames to display and 60 windows to display in. You paint most frames twice, but that gets you only to 50 fps. You also have to paint some frames three times. To fill the whole second.

That's why some videos are smooth. You record at 30 fps, edit at 30 fps timeline and then serve 30 fps video to 60 fps screen - every frame is painted exactly twice. Its consistent and stutter free.

If you record at anything other than 30 or 60 (or another value that you can divide 60 with without remainder), edit your video timeline at other than 30 or 60 fps, on 60 Hz screen (vaaaaast majority) you will always see stuttering.

People are arguing about the 24 vs 25 vs 30 fps while thinking about how much motion blur it allows them to capture, while completely ignoring the question "And what medium will I be watching the final output on?". Sure, if you are filming for a cinema, and you know the projector will run at 24 fps, feel free to record in 24 fps and enjoy stutter free motion blur. But if your goal is to view the video on a PC screen (or TV), make yourself a favour and record at 30 fps, and edit on 30 fps timeline.

Hi MaximusCZ

Thank you for this full explanation,

You just saved a cancellation request

 

[Cymru]

A lot of this stutter seen may well be the fact that a f/1.7 lens in bright sunlight with no ND filters is resulting in shutter speeds in excess of 1/1000th of a second.
I dont think any of the early-test groups had ND filters at all.

That and as above, frame rate conversions and all kinds of things for YouTube.
One thing i can see in sample videos that worries me slightly is the amount of pulsing/interframe flicker. Lots of anything with foliage or such close foreground looks to pulse. This is an issue the Mavic 1 suffered from badly.
I was hoping 150mbps/h265/8 bit would mean they didn't need to use interframe.

Ultimately though we need actual videos from actual users to download and play with rather then rely on YT to judge quality. Hopefully once real users start receiving hardware this week a lot of the raw videos will end up online.

EDIT:- The few tests ive found online seem to disagree with the 150Mbps video bit rates claimed by DJI.
From "Pilot Institute" YouTube (source:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

)where they've measured we have:

4k: 93Mbps
2.7k/60: 93 Mbps
2.7k: 52 Mbps
1080/60: 71Mbps
1080/30: 36Mbps

The few pieces of sample footage ive downloaded also show 4k as only having 93Mbps too.

Unless DJI fix this with a firmware upgrade then this lower bit rate is going to increase video compression and reduce quality.

 

[pilotinstitute]

EDIT:- The few tests ive found online seem to disagree with the 150Mbps video bit rates claimed by DJI.
From "Pilot Institute" YouTube (source:

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

)where they've measured we have:

4k: 93Mbps
2.7k/60: 93 Mbps
2.7k: 52 Mbps
1080/60: 71Mbps
1080/30: 36Mbps

The few pieces of sample footage ive downloaded also show 4k as only having 93Mbps too.

Unless DJI fix this with a firmware upgrade then this lower bit rate is going to increase video compression and reduce quality.

The unit we received wasn't able to record at 4k60fps so we are guessing that this is where the 150Mbps will be.

 

[EpicFlight]

I have no problems agreeing that properly judging video quality from any drone/camera requires access to the original footage. However, if the quality is good, and the author hasn't done errors or dumb choices in post, it is entirly possible to make the video look very good on YT as well.

Are you aware that YT does not simply store and stream what was uploaded, but recompresses the stored data, then changes resolution and recompresses AGAIN when streaming?

Evaluating anything technical about a YT stream is scrutinizing YT's image pipeline, little else. By the time you watch on YT, the video has been reprocessed several times.

And yes, stuttering and skipping can result from YT processing too.

 

[Cymru]

The unit we received wasn't able to record at 4k60fps so we are guessing that this is where the 150Mbps will be.

Makes sense but is still disappointing given compression artefacts at 93Mbps are going to be far more severe than 4k/30 at 150mbps.

It seems DJI enjoys adjusting the bit rate to give a fixed compression as opposed to a better quality amount of compression adjusted to match a fixed bit rate.

My M2P shows at times fairly noticeable compression artefacts at 4k/30 in both 8 and 10 bit at 100Mbps which i was hoping wouldn't be an issue on the Mini 3.

(On a side note, a good video. Actually giving proper figures and measurements that people can use on all aspects of the drone instead of "Oww this is good" type comments. Far more useful!)

 

[Deleted member 154956]

That's odd. It is stuttering very visibly for me.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

What do you see if you look at the above video from 43:06 to 43:35? That kind of shooting clearly brings out the worst conditions for stuttering, and for me it stutters like crazy. I wonder how much better it would have been at 4k60, but in that resolution ActiveTrack isn't available, so he couldn't have done that test anyways.

I guess ND filters will be a necessity, but it will perhaps take some time untl they become available.

Ok that one the stuttering was obvious, but that was your screen recording. That I would expect to see some transmission glitches to the remote, but the on-drone micro sd card recording should be fine. No?

 

[JoshC]

I was thinking the same when watching the the wind test DCrainmaker did where at 2:45 it's flying towards the lighthouse it didn't look smooth to me but I wondered if there's some transcoding issue as I've seen a similar effect recently when I uploaded a 5.7K60 video from a Virb 360 to YT.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

Isn't he recording off the controller there?

 

[Azunai2245]

Thought I'd throw my hat in the ring about my findings with my Mini 3 Pro, as I stumbled across this topic after looking at my first flight's footage. For context, I am a beginner drone pilot (this is my first drone). I like videography and photography as a hobby, so I have a basic working knowledge of the terminology. I definitely know what FPS is, but struggling to understand what setting is best to use in which scenario still, so that's why this is interesting for me.

I shot this footage in 30fps (bar the last shot of the field of mowed grass), and edited on a 30fps timeline in Premiere Pro. Other than cuts and arranging, I've made no changes to the clips. To my eyes, there's no difference between what I saw in the straight-out-of-drone video files, the timeline in Premiere, and the uploaded result in YouTube.

As you can see it's a bright sunny day. This was shot at around 10-11am. There's significant stutter in the panning shots, like at the beginning (0:08-0:12), mostly in the houses at the waterfront, when viewed in 4k. But it's just fine in the panning up shot at ~0:30 again. I see it mostly in fast pans, in grass/bushes and in the shots of houses.

Could this be fixed with an ND filter?

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

 

[scro]

Thought I'd throw my hat in the ring about my findings with my Mini 3 Pro, as I stumbled across this topic after looking at my first flight's footage. For context, I am a beginner drone pilot (this is my first drone). I like videography and photography as a hobby, so I have a basic working knowledge of the terminology. I definitely know what FPS is, but struggling to understand what setting is best to use in which scenario still, so that's why this is interesting for me.

I shot this footage in 30fps (bar the last shot of the field of mowed grass), and edited on a 30fps timeline in Premiere Pro. Other than cuts and arranging, I've made no changes to the clips. To my eyes, there's no difference between what I saw in the straight-out-of-drone video files, the timeline in Premiere, and the uploaded result in YouTube.

As you can see it's a bright sunny day. This was shot at around 10-11am. There's significant stutter in the panning shots, like at the beginning (0:08-0:12), mostly in the houses at the waterfront, when viewed in 4k. But it's just fine in the panning up shot at ~0:30 again. I see it mostly in fast pans, in grass/bushes and in the shots of houses.

Could this be fixed with an ND filter?

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

The stutter you're seeing is due to the bright lighting, leading to a fast shutter speed, which in turn eliminates most of the motion blur that would smooth this out. The target shutter speed for good motion blur at 30fps is 1/60s.

An appropriate density ND filter would fix this.

 

[Cymru]

The lack of ND filters and likely 1/4000th shutter speeds are going to make any rapid motion appear to stutter. This looks about right.
Also with rapid rotation or horizontal motion the codec struggles somewhat on all DJI drones due to the implementation - my Mavic 2 Pro you can only go VERY slowly parallel motion to avoid jarring stutter. The Mini 3 is somewhat better here as its less compressed.
Also remember a LOT of YouTube stuff loses quality due to frame rate conversion and transcoding.

 

[Azunai2245]

Cool - I wasn't worried as the footage without frantic movement was properly smooth, and had a hunch the bright light might be the cause of it, but nice to have my suspicions confirmed. Hopefully the above footage also helps OP in the evaluation

 

[Cymru]

You're going to need an ND64 or ND128 filter to bring the shutter speeds down to a level where you'll get smooth video.
*BUT* rapid rotations are always going to stutter to an extent due to the compression levels and settings DJI use.

 

[akdrone]

For example this one

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

Time 0:32 - 0:37, this shot is stuttering
Some of the shots are brilliant, but some gives a feeling of a laggy video
I have checked on different devices, stuttering remains

Perfectly smooth iPad Pro M1 12.9”

 

[Egeris]

Hello all! I have found this topic, because also own the Mini 3 pro and when I firstime saw the video from drone, there was this stuttering visible.

I was wondering what is wrong with my drone, because the Mavic 2 zoom (my second drone) never had this issue. So after some tests, I'm almost sure, there is nothing wrong with Mini 3 pro. The whole issue is somehow related to playback on my computers.

For example - this video

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

- when I play it on my work laptop, there is stutering really visible (as you desribed above). But the same video, which I play on my home computer, is nice smooth.

I did lot of tests on my own drone videos - with multiple player apps (VLC, PotPlayer, KMPlayer) and the stuttering effect is present, BUT in each player differently at different times. Sometimes it stutter for example from 0:02 to 0:06. When I close and open video again, it will stutter from 0:08 to 0:10 etc. So probably the issue is somewhere in player/codec.

Then I realize that I can switch rendering engine in player application. There is several options (DirectX 9, DirectX11 and other) - and each option acts differently. For example the playback on DirectX 11 is quite smooth in most cases - the stuttering occurs sometimes for short time.

So the question is how to deal with this issue? I need to edit my videos in DaVinci resolve, which is unfortunately stuttering. On other side, if some apps can play my videos smoothly, there is propably problem in something else.

The strange thing is that my Mavic 2 zoom never had this issue. Why? It is the same codec (h264)? Mybe there is different version/implementation of codec in mini 3 pro? I think that my PC is capable to play 4K videos (AMD Ryzen 3700X, Nvidia RTX 2070 super, 32 GB RAM...)...

And last question is - why is mini 3 pro recording only at 93 Mbps? This is really low bitrate, isn't it? For 10bit video... Mavic 2 zoom records at 100 Mbps...

 

[pilotinstitute]

And last question is - why is mini 3 pro recording only at 93 Mbps? This is really low bitrate, isn't it? For 10bit video... Mavic 2 zoom records at 100 Mbps...

The bitrate always depends on the image size. The higher bitrate that the manufacturer advertises is always for the highest frame rate at the highest resolution. Even the Mavic 2 Zoom doesn't record all resolutions at 100 Mbps.
The Mini 3 was 93MPbs at 4k30fps, but it is 150Mbps at 4k60.

 

[akdrone]

Hello all! I have found this topic, because also own the Mini 3 pro and when I firstime saw the video from drone, there was this stuttering visible.

I was wondering what is wrong with my drone, because the Mavic 2 zoom (my second drone) never had this issue. So after some tests, I'm almost sure, there is nothing wrong with Mini 3 pro. The whole issue is somehow related to playback on my computers.

For example - this video

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

- when I play it on my work laptop, there is stutering really visible (as you desribed above). But the same video, which I play on my home computer, is nice smooth.

I did lot of tests on my own drone videos - with multiple player apps (VLC, PotPlayer, KMPlayer) and the stuttering effect is present, BUT in each player differently at different times. Sometimes it stutter for example from 0:02 to 0:06. When I close and open video again, it will stutter from 0:08 to 0:10 etc. So probably the issue is somewhere in player/codec.

Then I realize that I can switch rendering engine in player application. There is several options (DirectX 9, DirectX11 and other) - and each option acts differently. For example the playback on DirectX 11 is quite smooth in most cases - the stuttering occurs sometimes for short time.

So the question is how to deal with this issue? I need to edit my videos in DaVinci resolve, which is unfortunately stuttering. On other side, if some apps can play my videos smoothly, there is propably problem in something else.

The strange thing is that my Mavic 2 zoom never had this issue. Why? It is the same codec (h264)? Mybe there is different version/implementation of codec in mini 3 pro? I think that my PC is capable to play 4K videos (AMD Ryzen 3700X, Nvidia RTX 2070 super, 32 GB RAM...)...

And last question is - why is mini 3 pro recording only at 93 Mbps? This is really low bitrate, isn't it? For 10bit video... Mavic 2 zoom records at 100 Mbps...

That was good of you to figure that out. Unfortunately it is not unusual for us to be looking for a solution in an area that seems to make sense but it’s completely off-base so kudos to you on that realization. My iMac will also not do a good job with 4K video and I’m considering a new computer also but my solution is to simply work with aliases which takes extra time in creating them but once they are all set the workflow is almost identical and completely smooth. I do have to re-render but for my short drone videos it’s not a big issue. This fall I will be doing a series of videos that will be much longer and so I may choose to get a new computer I’m sure Davinci resolve has a similar method.

 

[MaximusCZ]

when I play it on my work laptop, there is stutering really visible (as you desribed above). But the same video, which I play on my home computer, is nice smooth.

I pressume your Mini 3 is recording in h265, not h264. Your PC GPU has hardware to decode h265 on the fly, so you dont get stuttering there, but your laptop is probably older and doesnt come with hardware accelerated h265 decoder, forcing your CPU to decode the video in software. That is very intensive for 4k video, and your laptop cpu likely just cant keep up. Check your CPU utilization whrn stuttering is happening.

To fix this: get laptop with hardware h265 decoder, or switch your mini 3 to record in h264 instead.