Welcome Inspire Pilots!
Join our free DJI Inspire community today!
Sign up

X5S ProRes 422 HQ vs Mavic 3 ProRes 422 HQ (not good)

Joined
Jun 15, 2020
Messages
10
Reaction score
2
My work drone is the Inspire 2 with the X5S. Im taking off and landing on boats as most of my work is boating related. The reason I fly the Inspire is so I can shoot in ProRes 422 HQ - so with the release of the Mavic 3 Cine, I thought wow, what a great opportunity to loosen my load, I can get the same quality as the X5S in a smaller package and catching the mavic 3 is a lot easier than the inspire 2, especially on a boat over the water...

Well, I have downloaded many of sample files from the Mavic 3 Cine ProRes 422 HQ and from what I see, its no where comparable to the X5S ProRes 422 HQ. I pretty much only shoot in 2704 x 1520 | 50p on the X5S and comparing that to Mavic 3 Cine ProRes 422 HQ 4k or 5k the X5S looks so much better than the Mavic 3. When I compare the two, it makes the Mavic 3 - 422 HQ files look like h.264.

How is this possible when they both are shooting in ProRes 422 HQ, wouldn't you think the two images would look comparable? Its like the Mavic 3 is a fake ProRes 422 HQ.. The X5S bitrate is 900Mbps and the Mavic 3 Cine is 3,772 Mbps so whats going on here?


haha.png
 
Last edited:
  • Like
Reactions: Joan63
So I'll just share my thoughts on this. I do not know if this is true so this is just my opinion. While the Mavic 3 Cine is a way newer system I don't feel it was designed for a professional look. While it is an amazing drone that can shoot extremely high-quality content it has its drawbacks. For most people, you wouldn't notice a huge difference in the quality. For starters, the lens itself is a huge difference. The one on the drone is very limited in size and can't have as many internal lenses or corrections as let's say an Olympus. I'm not calling the Mavics bad by any means as it's incredible what they were able to do, but the size has to mean something here. Also from what I know, I thought the ProRes was still being worked on by DJI. I'm still not sure if it is fully ready software-wise to be a replacement yet for the Inspire Series. Also, I like the size of the Mavic more than the Inspire and the fact it is running all of the newer tech and features the Inspire doesn't have access to yet so that is a plus. I have a small boat so I know what a pain larger drones can be to launch and catch from one. For filming, I think the Inspire is still better like you are saying. It has been out for a while so all of the kinks of RAW video have been worked out unlike the Cine yet. It may take time for it to be worked on. It is technically a first-generation product because it is nothing like the Mavic 2 hardware-wise. I think that is why it looks so different. In my opinion, though I think it mainly is the fact that while the sensor size is the same, the actual camera system is vastly different. Also funny enough I would actually assume that the Inspire would be easier to capture video from because of the active track being a separate system from the drone's flight. So the operator can just fly while the camera does the work for you. I haven't personally flown either so again this is just my thoughts on the matter. Also, remember you are comparing stock footage to footage you personally took so the settings may have not been right on the Mavic.

Also, I love the edit you did! Looks like an interesting concept if that were to be real.
 
My work drone is the Inspire 2 with the X5S. Im taking off and landing on boats as most of my work is boating related. The reason I fly the Inspire is so I can shoot in ProRes 422 HQ - so with the release of the Mavic 3 Cine, I thought wow, what a great opportunity to loosen my load, I can get the same quality as the X5S in a smaller package and catching the mavic 3 is a lot easier than the inspire 2, especially on a boat over the water...

Well, I have downloaded many of sample files from the Mavic 3 Cine ProRes 422 HQ and from what I see, its no where comparable to the X5S ProRes 422 HQ. I pretty much only shoot in 2704 x 1520 | 50p on the X5S and comparing that to Mavic 3 Cine ProRes 422 HQ 4k or 5k the X5S looks so much better than the Mavic 3. When I compare the two, it makes the Mavic 3 - 422 HQ files look like h.264.

How is this possible when they both are shooting in ProRes 422 HQ, wouldn't you think the two images would look comparable? Its like the Mavic 3 is a fake ProRes 422 HQ.. The X5S bitrate is 900Mbps and the Mavic 3 Cine is 3,772 Mbps so whats going on here?


View attachment 32014
Must be the inferior lens of the Mavic 3. Glass makes a difference.
 
I've got an Inspire2/X5s and wondering why I'm so impressed at first by the Mavic 3 Pro Cine footage (the gimbal can stay centered and not twitch too much?) ... until I look a little closer ..

I don't know how to say this, or to be able to prove this, but I suspect that the "ProRes" part of this equation is only that Mavic 3 is basically shooting h265 or something equivalent(low bitrate) then re-encoding it in ProRes on the fly ... so the "test" if I can be scientific here (I can't) is to shoot h.265 and also ProRes on the Mavic, and compare the two. If DJI were smart and clever, they would downgrade the h.265 so it always looks worse ...

I'm looking not at the lens, but at the internal compression. I see weird flickerings and "fake" detail enhancements where there should be none. Yes, I'm looking at 400% magnification and larger, but if it's really just a lossy codec cased in a "pro" intermediate, then it makes more sense to A) don't bother getting the Cine version which is a slow waste of money and time and B) if you already did, just shoot in h.265 and transcode. Tests will happen.

Again, I can't say for certain, but I'm seeing flickering and artifacts where there should be none. (Comparing to ProRes shot natively on X-H2s, X5s, ZCAM E2-M4).

Other than image complaints (expectations dashed) and some gimbal twitchiness at 166mm, the bird is a delight.
 
I was playing with M3 Cine ProRes RAW footage and quickly found it’s not behaving like real raw but more interesting was that the corrections of the raw I made in davinci work without any significant difference to the H264 too. Then I quickly understood this drone isn’t for me and all the hype died for me. Can’t believe they sell this fake raw for 5 grants.
 
I was playing with M3 Cine ProRes RAW footage and quickly found it’s not behaving like real raw but more interesting was that the corrections of the raw I made in davinci work without any significant difference to the H264 too. Then I quickly understood this drone isn’t for me and all the hype died for me. Can’t believe they sell this fake raw for 5 grants.
Unless I'm mistaken (I have been before, and lots) this is NOT ProRes RAW ... nor was it purported to be. That's different. It should not be expected to perform like that. However, in my initial tests today (specifically shooting the same scene/movement with only changing codec) I can see very very little difference between h.265 and Prores 422. (it's debatable whether 422 and HQ are visually different and yes I'm one who looks at files at 400% mag and up)

I think the bird is quite worthy. But it's not necessarily worth the money for Cine version in my opinion. Not even close.
 
I've got an Inspire2/X5s and wondering why I'm so impressed at first by the Mavic 3 Pro Cine footage (the gimbal can stay centered and not twitch too much?) ... until I look a little closer ..

I don't know how to say this, or to be able to prove this, but I suspect that the "ProRes" part of this equation is only that Mavic 3 is basically shooting h265 or something equivalent(low bitrate) then re-encoding it in ProRes on the fly ... so the "test" if I can be scientific here (I can't) is to shoot h.265 and also ProRes on the Mavic, and compare the two. If DJI were smart and clever, they would downgrade the h.265 so it always looks worse ...

I'm looking not at the lens, but at the internal compression. I see weird flickerings and "fake" detail enhancements where there should be none. Yes, I'm looking at 400% magnification and larger, but if it's really just a lossy codec cased in a "pro" intermediate, then it makes more sense to A) don't bother getting the Cine version which is a slow waste of money and time and B) if you already did, just shoot in h.265 and transcode. Tests will happen.

Again, I can't say for certain, but I'm seeing flickering and artifacts where there should be none. (Comparing to ProRes shot natively on X-H2s, X5s, ZCAM E2-M4).

Other than image complaints (expectations dashed) and some gimbal twitchiness at 166mm, the bird is a delight.
Are you using d-log? D-log is less processed than the other color profiles. Also keep in mind only HLG and d-log are 10 bit
 
Last edited:
this is NOT ProRes RAW
Oh yes sorry it’s ProRes HQ and I was comparing it with raw from inspire. But the quality of the ProRes was not much different than the h264 and both cannot stand against raw footage. So it’s not even close to “Cine”. They had to name it Mavic 3 TV :D
 
TL/DR you need to shoot in d-log. You’ll need to do some extra processing to deal with noise and sharpness issues created by the poor quality lens but it doesn’t have the artifacts you are talking about.

ProRes is just the format the video comes in, it does not change the visual look of the video and is no indication of quality. 200 Mbps h.265 from the regular Mavic 3 is already a high enough bit rate that you should not visually see compression artifacts in 4k or even 5k unless you are zoomed in 400% which, I mean this in the most jolly and inoffensive way possible, you must already know is crazy right? Therefore expecting to see similar image quality of ProRes from different cameras and lenses is like expecting the quality of h.264 of different cameras to look the same.

While it’s not like the Inspire 2 that has two completely separate image pipelines that work simultaneously, the Mavic 3 does have two different imaging pipelines. One for D-log and one for everything else. Lots of processing is done on the “everything else” pipeline, no matter the codec you use. I find it likely this is the cause of artifacts you are seeing. However, the D-log is fairly unprocessed and doesn’t have those artifacts you are talking about.

Also, unlike other cameras the Mavic 3 only records 10 bit video when using the D-log or HLG. Any other color profile is crap.

While I find the X5s has much better color science and the quality of the lenses are obviously better, I have been impressed by the latitude and malleability of the D-log even in H.265 on the Mavic 3.
Oh, I'm impressed too. It's just that I don't see a huge difference in the h.265 after transcoding. I know what you think I mean: That the Prores should "look different". I know it's just a container .. it's containing a low bitrate stream in this case. I've been shooting various codecs for years. I know what I mean .. for you to know that, I'd have to clarify: What I mean here, is that due to the low bitrate pipeline in the M3, the ProRes is not really worth the price. I'm not necessarily saying it's a bad bird/camera, but the Cine version is not living up to the hype. Buy cheaper, shoot longer, transfer quicker, and transcode only the good bits. With the same or close results. Close enough to not matter, unless pixel peeping. Content matters more than the small difference I'm realising with the ProRes. Prores should mean less compression (not none). So less loss of detail. But the detail is not there in the first place! Definitely better than the Mavic 2 Pro. Not as good as the Inspire 2/X5s. Who would think that? Well, I did, at first, then quickly got that bubble burst, even on the main camera, in DLOG, and no, not just comparing the 5.2K C-DNG .. but also the ProRes.

I'm also torture-testing the cameras. Complicated detail-intense busy scenes .. you know .. trees and forests and extreme dynamic range (dark forest/bright clouds) coupled with movement. That "smearing" in both ProRes and h.265 suggests low internal bitrate. In some cameras, you see the smearing happen with h.264/265, sooner with the lower bitrate options.

I'm comparing ProRes vs h.265 also on the Fuji H2s as well as the ZCAM E2-M4. On those cameras it's not hugely different either .. on the Fuji it's 360 or 720Mb/s, by the way .. the 360 DOES work on the SD card, (which is surprising) but the 700+ requires the CF Express. The micro SD cards on the Mavic can't deal with that, so 200Mbps it is.

Would be super nice of DJI to provide a 720 Mb/s h.265 for those with the onboard SSD, but they do not.

When I say "looks different" I also mean I'm able to aggressively push and pull at the 200mbps as hard as I can the Prores, with the same results. There are artifacts, and smearing, on the h.265, but not much less on the ProRes ... and it's not crazy at all to look at footage at 400+ percent. Sometimes it's the small things that matter. They add up. I've been doing it for years. I've seen things that people don't understand .. I try to rub their noses in it, but they can't or won't see it. It's not that it's not real .. it's that it isn't important if you're doing it right, but I often am looking for camera solutions that help me find joy when I get back and despite doing it wrong, find good imagery from unscripted shoots in wild conditions. Yeah .. nature.

Also, and most important here, is that we can't shoot DLOG (or 10-bit?) using the other cameras. Part of the value of this bird are the longer lenses. For what they are, they are amazing. Not being able to shoot in HLG or DLOG on them is annoying.
 
Oh yes sorry it’s ProRes HQ and I was comparing it with raw from inspire. But the quality of the ProRes was not much different than the h264 and both cannot stand against raw footage. So it’s not even close to “Cine”. They had to name it Mavic 3 TV :D
Hilarious!
 
Oh, I'm impressed too. It's just that I don't see a huge difference in the h.265 after transcoding. I know what you think I mean: That the Prores should "look different". I know it's just a container .. it's containing a low bitrate stream in this case. I've been shooting various codecs for years. I know what I mean .. for you to know that, I'd have to clarify: What I mean here, is that due to the low bitrate pipeline in the M3, the ProRes is not really worth the price. I'm not necessarily saying it's a bad bird/camera, but the Cine version is not living up to the hype. Buy cheaper, shoot longer, transfer quicker, and transcode only the good bits. With the same or close results. Close enough to not matter, unless pixel peeping. Content matters more than the small difference I'm realising with the ProRes. Prores should mean less compression (not none). So less loss of detail. But the detail is not there in the first place! Definitely better than the Mavic 2 Pro. Not as good as the Inspire 2/X5s. Who would think that? Well, I did, at first, then quickly got that bubble burst, even on the main camera, in DLOG, and no, not just comparing the 5.2K C-DNG .. but also the ProRes.

I'm also torture-testing the cameras. Complicated detail-intense busy scenes .. you know .. trees and forests and extreme dynamic range (dark forest/bright clouds) coupled with movement. That "smearing" in both ProRes and h.265 suggests low internal bitrate. In some cameras, you see the smearing happen with h.264/265, sooner with the lower bitrate options.

I'm comparing ProRes vs h.265 also on the Fuji H2s as well as the ZCAM E2-M4. On those cameras it's not hugely different either .. on the Fuji it's 360 or 720Mb/s, by the way .. the 360 DOES work on the SD card, (which is surprising) but the 700+ requires the CF Express. The micro SD cards on the Mavic can't deal with that, so 200Mbps it is.

Would be super nice of DJI to provide a 720 Mb/s h.265 for those with the onboard SSD, but they do not.

When I say "looks different" I also mean I'm able to aggressively push and pull at the 200mbps as hard as I can the Prores, with the same results. There are artifacts, and smearing, on the h.265, but not much less on the ProRes ... and it's not crazy at all to look at footage at 400+ percent. Sometimes it's the small things that matter. They add up. I've been doing it for years. I've seen things that people don't understand .. I try to rub their noses in it, but they can't or won't see it. It's not that it's not real .. it's that it isn't important if you're doing it right, but I often am looking for camera solutions that help me find joy when I get back and despite doing it wrong, find good imagery from unscripted shoots in wild conditions. Yeah .. nature.

Also, and most important here, is that we can't shoot DLOG (or 10-bit?) using the other cameras. Part of the value of this bird are the longer lenses. For what they are, they are amazing. Not being able to shoot in HLG or DLOG on them is annoying.
Yea sorry I actually previously pared down my comment because it sounded condescending on second read which wasn’t my intention.

I agree with your assessment that shooting straight to ProRes 422 HQ 10 bit over h.265 420 10 bit isn’t worth it with from a visual quality difference or even grading standpoint with the important caveat of using visual effects. I do high end commercial real estate videos and use 3D camera tracking to add 3D overlays on to drone footage and I get much better tracking data using ProRes due to the intra-frame compression which can be the difference between a shot working or not working so in my case it’s worth it.

I remain skeptical that they first compress it to a low bit rate and then transcode it to ProRes. It just seems like that would take more processing power and more memory bandwidth to do then encode it to ProRes directly. While bit rate is a factor in ability to push and pull I just don’t think it has the dynamic range to really make use of all that bandwidth.

One correction, on the Mavic 3 Pro Cine can record 10 bit normal color on all 3 cameras in ProRes only.
 
Last edited:
Ooh ... thanks for correcting the thing about 10-bit on all three cams ... I was confused for a second, as I thought it was. In fact, I found the big bugbear to be the "normal" on the 166mm, which is waaaay hyped .. happily I brought it down to an equivalent look to the others fairly easily, so they all kinda match. In good light and/or with good subject (so one is not looking at the quality, but rather at the content) it's great. I really love telephoto in the air.

No worries ... I was not offended, but sought to clarify. It's really hard to know how to talk to different people online ... not really knowing where each person's knowledge level or understanding is at is challenging. I'm probably smack dab in the "I know enough to be dangerous, but not enough to transcend the tech" level. : ) It's a boon to have patient, more skilled/knowledgeable people taking the time, so thank you.

Agreed on the tracking ... even simple tracked masks are difficult ... and yes, even manual roto work is a pain if interpolated pixels are blobbing all over the place. (roto work is another reason to mag to 400% and more sometimes) I'll keep that in mind when shooting shots I know are likely to be tracked or such.

No, I wasn't suggesting they are encoding h.265 and then transcoding ... it is just that I suspect the data stream from sensor to processor is fairly low, lower than is should be, before being encapsulated as ProRes.

Were I to advise someone, I'd say get the non-Cine version unless they specifically need a tad better chance to track/roto, etc. Had DJI made the SSD removeable I'd be more inclined to say get it if you can afford it, but having to connect the bird physically is a bummer, and the bigger file sizes with no massive gain is a bummer. I like to swap cards before each flight ... things can happen ... so that now means a data dump between each flight .. while the light fades. Phooey to that!

------------------------


Previously writ: (I had responded to it, but then found it gone, so this is what I was responding to!)


Yea sorry I actually previously pared down my comment because it sounded condescending on second read which wasn’t my intention.

I agree with your assessment that shooting straight to ProRes 422 HQ 10 bit over h.265 422 10 bit isn’t worth it with from a visual quality difference or even grading standpoint with the important caveat of using visual effects. I do high end commercial real estate videos and use 3D camera tracking to add 3D overlays on to drone footage and I get much better tracking data using ProRes due to the intra-frame compression which can be the difference between a shot working or not working so in my case it’s worth it.

I remain skeptical that they first compress it to a low bit rate and then transcode it to ProRes. It just seems like that would take more processing power and more memory bandwidth to do then encode it to ProRes directly. While bit rate is a factor in ability to push and pull I just don’t think it has the dynamic range to really make use of all that bandwidth.

One correction, on the Mavic 3 Pro Cine can record 10 bit normal color on all 3 cameras.
 

New Posts

Members online

Forum statistics

Threads
22,334
Messages
210,842
Members
34,935
Latest member
liamjames