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

First Night Shoot and Client Isn't Happy With Footage

Joined
Jun 21, 2019
Messages
6
Reaction score
1
Age
33
Location
Brooklyn, NY
Website
www.atmosphericarts.com
I've been working with a local pro-sports team to document the construction of their new stadium, and I recently just did a shoot of the building about an hour after sunset. While this was my first commercial night flight ever (I do have the waiver!), I consider myself pretty knowledgeable about how cameras and camera settings work so I figured I knew the right settings for the shot (DLog, wide-open aperture, 180* shutter, ProRes 422HQ). I typically have the MicroSD card set to capture 4k Rec.709 H.265 clips along with the ProRes. After reviewing the footage, I noticed that the H.265 footage looked quite nice and with minimal noise, yet the ProRes/Log clip was extremely noisy but definitely had more shadow detail. From my understanding of how EI mode works (and how the H.265 versions work), the video is always captured at native ISO (400) and when I bumped it up to 1600 ISO it is just effectively applying a LUT to bump the gain by a couple stops (reversible in post). I also assumed that the H.265 version may have had less noise because DJI is applying well-tuned noise reduction to it along with the Rec.709 color profile.

That said, I figured my client (who is admittedly way more knowledge about about video than me) would know why the ProRes looked so much noisier, and he could apply his own noise reduction and stop it down a bit if it was too bright. Unfortunately I just got the feedback that he considers the ProRes footage unusable due to noise and the H.265 versions passable. How is that possible that DJI's noise reduction is so well tuned that a video professional prefers the H.265 version of a clip to a LOG ProRes file? Is my understanding wrong here? Any advise on Noise Reduction settings in Premiere for X5S DLOG low-light footage?

I can share a few of the clips privately if you're curious but I'd prefer not post them here. Thanks for any help you can provide!
 
Last edited:
  • Like
Reactions: Donnie Frank
Joined
Jun 28, 2016
Messages
220
Reaction score
82
Age
58
Watching this thread with great interest....

Like you, atmosphericarts, I'm fairly versed in photography/cinematography, but am not a schooled expert. So, not to muddy this thread, but I have to ask.

This sentence confused me a bit:

"(DLog, wide-open aperture, 180* shutter, ProRes 422HQ). I typically have the MicroSD card set to capture 4k Rec.709 H.265 clips along with the ProRes."

If I'm reading this correctly, you're applying a Rec.709 LUT to the h.265 MicroSD footage AND the ProRes footage, yes? I'm going to guess "no," but your syntax lead me to believe that MAYBE you were applying the Rec.709 LUT to the ProRes footage. If so, I wouldn't mind diving a little deeper into your settings.

I did some night shooting a couple years ago using an X5, D-Log, 180° shutter, 30fps, h.264, WB 4200K, that turned out very nice. The sweet spot for night shooting was ISO 800. 1600 got too grainy. The client was very pleased with the footage, which was handed off to a professional production company for grading and editing.

Here's the footage:


It's easy to spot the lower bitrate in the details, but the color and light look good IMHO.

I look forward to seeing how this thread plays out.

D
 
  • Like
Reactions: atmosphericarts
Joined
Jun 21, 2019
Messages
6
Reaction score
1
Age
33
Location
Brooklyn, NY
Website
www.atmosphericarts.com
Watching this thread with great interest....

Like you, atmosphericarts, I'm fairly versed in photography/cinematography, but am not a schooled expert. So, not to muddy this thread, but I have to ask.

This sentence confused me a bit:

"(DLog, wide-open aperture, 180* shutter, ProRes 422HQ). I typically have the MicroSD card set to capture 4k Rec.709 H.265 clips along with the ProRes."

If I'm reading this correctly, you're applying a Rec.709 LUT to the h.265 MicroSD footage AND the ProRes footage, yes? I'm going to guess "no," but your syntax lead me to believe that MAYBE you were applying the Rec.709 LUT to the ProRes footage. If so, I wouldn't mind diving a little deeper into your settings.

I did some night shooting a couple years ago using an X5, D-Log, 180° shutter, 30fps, h.264, WB 4200K, that turned out very nice. The sweet spot for night shooting was ISO 800. 1600 got too grainy. The client was very pleased with the footage, which was handed off to a professional production company for grading and editing.

Here's the footage:


It's easy to spot the lower bitrate in the details, but the color and light look good IMHO.

I look forward to seeing how this thread plays out.

D
I have the SSD recording in DLog (and would normally apply a DLOG>Rec.709 LUT in post). I have the MicroSD card set up to record H.265 directly in Rec.709 (no LUTs in Post if we end up using these) so we can quickly get a feel for the final look after export, and also so the monitor looks closer to life. I was shooting at ISO 1600 but was under the impression that in EI mode everything can be reversed in post by just lowering the gain.
 
Joined
Jul 6, 2014
Messages
314
Reaction score
147
EI mode is very poorly documented.

ISO 1600 will produce noisy footage on any camera. What brought you to ISO 1600? What frame ratter and shutter speed were you using? What lens and aperture?

My understanding of photography and technology is pretty good, and I don't think that the MicroSD footage can possibly be anything but that same ProRes footage with noise reduction applied.

But I am not going to pretend that I know more than the professional who hired you. I am also interested in seeing the two different clips. And a third clip: the ProRes footage with some noise reduction applied.
 
Joined
Jul 6, 2014
Messages
314
Reaction score
147
PS I have used EI mode extensively on film sets, but always during the daytime, and always at ISO 100. We use ND filters and narrow apertures to get the appropriate shutter speed. The producers I worked for were very pleased with the resulting footage, given that they didn't have to pay someone to fly an Arri, which they used for all other shots.
 
Joined
Oct 14, 2015
Messages
17
Reaction score
8
Age
60
Try dropping your ISO to 100. If you are going to shoot at night then you need to be shooting with a Sony A-7.
 
  • Like
Reactions: AP3
Joined
Jun 21, 2019
Messages
6
Reaction score
1
Age
33
Location
Brooklyn, NY
Website
www.atmosphericarts.com
EI mode is very poorly documented.

ISO 1600 will produce noisy footage on any camera. What brought you to ISO 1600? What frame ratter and shutter speed were you using? What lens and aperture?

My understanding of photography and technology is pretty good, and I don't think that the MicroSD footage can possibly be anything but that same ProRes footage with noise reduction applied.

But I am not going to pretend that I know more than the professional who hired you. I am also interested in seeing the two different clips. And a third clip: the ProRes footage with some noise reduction applied.
I shoot in EI mode with the I2 99% of the time (typically at ISO 100), but this was a particularly dark shoot due to it being a poorly lit (on the exterior at least) construction site. ISO 1600 looked best on the histogram and monitor but it definitely seems to be extremely noisy compared to 800 and 400, even after increasing the exposure to match in post. So I guess the lesson learned is to either shoot in cDNG or expose to the left at night.
 
Joined
Jun 28, 2016
Messages
220
Reaction score
82
Age
58
I shoot in EI mode with the I2 99% of the time (typically at ISO 100), but this was a particularly dark shoot due to it being a poorly lit (on the exterior at least) construction site. ISO 1600 looked best on the histogram and monitor but it definitely seems to be extremely noisy compared to 800 and 400, even after increasing the exposure to match in post. So I guess the lesson learned is to either shoot in cDNG or expose to the left at night.
I would also recommend the use of ISO800.

Before I did my night shooting for this company, I did a bunch of night time test flights. The Inspire 1 produces an .srt file (subtitle file), which allows me to view camera settings in real time. With this, I was able to figure out which settings worked best for night shooting (settings shared in previous post). This was using the X5 camera with stock lens. Obviously the histogram was not ideal (underexposed), but footage was "light enough" that it could be pushed in post.

My personal experience has been that D-Log produces the greatest dynamic range, with "Cinelike" coming in a close second. Obviously, for night shooting you want the widest dynamic range you can squeeze out of your camera, so I think you would find your night video easier to work with in post if you shot D-Log. If you conduct some night time test flights you'll see what I mean.

NOTE: I believe the newer DJI offerings no longer create .srt files, but instead imbed the "subtitle" in the video file. I don't like this "new way," but I guess it is what it is. At any rate, it's a great tool for figuring out which settings work best for a given situation.

D
 
Joined
Apr 23, 2020
Messages
22
Reaction score
6
Location
Salt Lake City, UT
I've been working with a local pro-sports team to document the construction of their new stadium, and I recently just did a shoot of the building about an hour after sunset. While this was my first commercial night flight ever (I do have the waiver!), I consider myself pretty knowledgeable about how cameras and camera settings work so I figured I knew the right settings for the shot (DLog, wide-open aperture, 180* shutter, ProRes 422HQ). I typically have the MicroSD card set to capture 4k Rec.709 H.265 clips along with the ProRes. After reviewing the footage, I noticed that the H.265 footage looked quite nice and with minimal noise, yet the ProRes/Log clip was extremely noisy but definitely had more shadow detail. From my understanding of how EI mode works (and how the H.265 versions work), the video is always captured at native ISO (400) and when I bumped it up to 1600 ISO it is just effectively applying a LUT to bump the gain by a couple stops (reversible in post). I also assumed that the H.265 version may have had less noise because DJI is applying well-tuned noise reduction to it along with the Rec.709 color profile.

That said, I figured my client (who is admittedly way more knowledge about about video than me) would know why the ProRes looked so much noisier, and he could apply his own noise reduction and stop it down a bit if it was too bright. Unfortunately I just got the feedback that he considers the ProRes footage unusable due to noise and the H.265 versions passable. How is that possible that DJI's noise reduction is so well tuned that a video professional prefers the H.265 version of a clip to a LOG ProRes file? Is my understanding wrong here? Any advise on Noise Reduction settings in Premiere for X5S DLOG low-light footage?

I can share a few of the clips privately if you're curious but I'd prefer not post them here. Thanks for any help you can provide!
I did some tests today and there’s but a minute difference between the total amount of noise in EI mode ProRes DLOG vs it’s rec 709 H265 counterpart. This is using Neat Video to build a noise profile which gives me an objective measure of noise.

100 EI gives you a fairly even spread of noise over the luminance range where as 1600 EI give you high noise in the shadows but very low noise in the highlights, even less than at 100 EI. 100 EI overall gives you less noise overall even if you have to boost the shadows in post. The amount of shadow recovery you can get at 100 EI (edit: from the ProRes DLOG) is pretty extraordinary. Surprised me quite a bit.
 
Last edited:
  • Like
Reactions: Donnie Frank
Joined
Nov 17, 2017
Messages
145
Reaction score
57
Location
Australia & Europe
I did some tests today and there’s but a minute difference between the total amount of noise in EI mode ProRes DLOG vs it’s rec 709 H265 counterpart. This is using Neat Video to build a noise profile which gives me an objective measure of noise.

100 EI gives you a fairly even spread of noise over the luminance range where as 1600 EI give you high noise in the shadows but very low noise in the highlights, even less than at 100 EI. 100 EI overall gives you less noise overall even if you have to boost the shadows in post. The amount of shadow recovery you can get at 100 EI is pretty extraordinary. Surprised me quite a bit.
I need to start with declaring that I am not a video expert and my knowledge of the intricate realm of video is limited.
But I have been a pro architectural photographer for over 20 years. Shooting RAW often in very challenging light. Processing in CaptureOne as my primary raw converter and then using PS CC to finetune.Though my knowledge is not specifically video I know thing or two about how to process, edit and adjust photos.
That was a long and boring intro and now back to my point, finally....
At one point of time I was considering upgrading my I2 to ProRes licence and buying the expensive media storage sticks, reader etc... Before I pulled the trigger I asked one of my pro ProRes shooter friend to share with me some of his footage, shot during the day. I spent considerable amount of time in Premiere Pro CC analyzing and comparing his ProRes footage with some of my own clips shot in .h265 with D-Cinelike profile.
Even though I could see some difference, especially in deep shadows where the ProRes retained more color info and it was cleaner, the real difference in overal quality, especially when the final video was exported into MP4 and viewed on 4K TV set, or worse exported to Vimeo or YT and watched there, was so negligible that I abandoned my idea. Even though I am not a pro video shooter I do understand that in the high end video production ProRes or now ProResRaw is a gospel. I only shoot video with my x5s mostly for fun doing my own little personal projects - cinematic landscapes. Only occasionally I have been asked to shoot video for a client but I have so far had zero complaints. Maybe my clients are not too fussy?‍♂️. But I only shoot during day or just before sunrise and after sunset and always make sure to expose properly and then tweek to best of my ability (applying my photo skills) in post. I find that .h265 (or even .264 for that matter) clips are reasonably flexible in post, relatively noise free and take well highlights and shadows recovery (to a point), curves and WB adjustments or even color grading for artistic licence if desirable.
That's just 2c worth from a video layman
 
Last edited:
Joined
Apr 23, 2020
Messages
22
Reaction score
6
Location
Salt Lake City, UT
I need to start with declaring that I am not a video expert and my knowledge of the intricate realm of video is limited.
But I have been a pro architectural photographer for over 20 years. Shooting RAW often in very challenging light. Processing in CaptureOne as my primary raw converter and then using PS CC to finetune.Though my knowledge is not specifically video I know thing or two about how to process, edit and adjust photos.
That was a long and boring intro and now back to my point, finally....
At one point of time I was considering upgrading my I2 to ProRes licence and buying the expensive media storage sticks, reader etc... Before I pulled the trigger I asked one of my pro ProRes shooter friend to share with me some of his footage, shot during the day. I spent considerable amount of time in Premiere Pro CC analyzing and comparing his ProRes footage with some of my own clips shot in .h265 with D-Cinelike profile.
Even though I could see some difference, especially in deep shadows where the ProRes retained more color info and it was cleaner, the real difference in overal quality, especially when the final video was exported into MP4 and viewed on 4K TV set, or worse exported to Vimeo or YT and watched there, was so negligible that I abandoned my idea. Even though I am not a pro video shooter I do understand that in the high end video production ProRes or now ProResRaw is a gospel. I only shoot video with my x5s mostly for fun doing my own little personal projects - cinematic landscapes. Only occasionally I have been asked to shoot video for a client but I have so far had zero complaints. Maybe my clients are not too fussy?‍♂️. But I only shoot during day or just before sunrise and after sunset and always make sure to expose properly and then tweek to best of my ability (applying my photo skills) in post. I find that .h265 (or even .264 for that matter) clips are reasonably flexible in post, relatively noise free and take well highlights and shadows recovery (to a point), curves and WB adjustments or even color grading for artistic licence if desirable.
That's just 2c worth from a video layman
To be clear that wasn’t the point I was trying to make and I’ve updated my last post to try and clarify that. The ISO 100 h.265 was hopelessly under exposed in the shadows and could not be recovered. The ProRes DLOG EI 100 was what surprised me as far as being able to recover shadows.

One might expect the noise to actually be somewhat better from the h.264/h.265 due to the the possibility of in camera noise reduction but I didn’t find evidence that in camera noise reduction was applied on the H265 and not the ProRes. I may do the same experiment with ProRes and cDNG to see if there is noise reduction added to the ProRes, though, I would expect the RAW to look much different straight out of camera so I may have to convert the RAW to DLOG first.

For me personally cDNG/RAW is the Gospel. Not that I do anything that is “needs” to be in RAW I just find it much easier to work with than even ProRes. Takes 1/4 the time to correct RAW than DLOG or H265.
 
Joined
Nov 17, 2017
Messages
145
Reaction score
57
Location
Australia & Europe
To be clear that wasn’t the point I was trying to make and I’ve updated my last post to try and clarify that. The ISO 100 h.265 was hopelessly under exposed in the shadows and could not be recovered. The ProRes DLOG EI 100 was what surprised me as far as being able to recover shadows.

One might expect the noise to actually be somewhat better from the h.264/h.265 due to the the possibility of in camera noise reduction but I didn’t find evidence that in camera noise reduction was applied on the H265 and not the ProRes. I may do the same experiment with ProRes and cDNG to see if there is noise reduction added to the ProRes, though, I would expect the RAW to look much different straight out of camera so I may have to convert the RAW to DLOG first.

For me personally cDNG/RAW is the Gospel. Not that I do anything that is “needs” to be in RAW I just find it much easier to work with than even ProRes. Takes 1/4 the time to correct RAW than DLOG or H265.
To be clear I did not post to contradict your findings. I simply wanted to share my experience and a very limited one at that. But to me the .h265 footage if properly exposed and WB correctly set in GO4 prior to shooting can, at least to me, deliver comparable image quality to ProRes footage. That is when looking at the final video in MP4 format delivered or uploaded to Vimeo or YT. That was my point and that was quite surprising to me. I was expecting the ProRes would blow my .h265 footage out of the water but to me at least it did not. It might be a different story if the original clips were heavily underexposed or way off the correct WB or perhaps shot at night. Then the ProRes might shine more and no doubt the Cinema DNG would be the king of all, though at the considerable toll of having to deal with huge files and needing a powerful comp for post.
 
Joined
Apr 23, 2020
Messages
22
Reaction score
6
Location
Salt Lake City, UT
To be clear I did not post to contradict your findings. I simply wanted to share my experience and a very limited one at that. But to me the .h265 footage if properly exposed and WB correctly set in GO4 prior to shooting can, at least to me, deliver comparable image quality to ProRes footage. That is when looking at the final video in MP4 format delivered or uploaded to Vimeo or YT. That was my point and that was quite surprising to me. I was expecting the ProRes would blow my .h265 footage out of the water but to me at least it did not. It might be a different story if the original clips were heavily underexposed or way off the correct WB or perhaps shot at night. Then the ProRes might shine more and no doubt the Cinema DNG would be the king of all, though at the considerable toll of having to deal with huge files and needing a powerful comp for post.
I understand. Everybody has different needs. If shooting in full light, properly exposed, have enough dynamic range, don’t need to match the footage with another camera, working in rec. 709, and don’t need to transcode it for VFX then there probably isn’t any benefit to ProRes or cDNG.
 
Joined
Apr 22, 2016
Messages
116
Reaction score
24
Location
Sofia
Website
goldenskybirds.com
In my experience shooting at night with x5's with more than ISO 400 is not acceptable by some producers because of the noise and image degradation and the footage cannot be matched to the rest of the cameras used in the shoot. Our settings for night shooting are usually - f1.7, shutter 1/25, ISO 400, cDNG, fps 23.976. The footage still have to be post-denoised with neat video and gained so to match with the RED's or ARRI's. And sometimes we set the shutter to 1/50. We use ISO 800 only in extremely dark situations, where there are no buildings and street lights around to cast some light like you are in the middle of a desert, but after denoising the image is so degraded that it may not pass QC especially if had to be matched to other camera.

X7 is an option for I2 and it will be better for night scenes, but the camera and all lenses are serious investment. Also for this type of jobs you may consider in another heavy lift drone with Sony A7 or RED Gemini which is also e serious investment but at least you can change the camera with wathever you like and not rely on all that only to DJI. We have M600Pro and are renting RED camera to complete the setup and fly it for the guys who really need it.
 

Members online

Forum statistics

Threads
2,333
Messages
24,198
Members
4,413
Latest member
Tom C