Color mode & Dynamic Range

[Raptorman0909]

Raptorman, there is After Effects for that kind of things.

After Effects has a lot of stuff and there are tracking features built into it, but does it: permit accurate tracking of the sky even as the camera turns changing the amount of the FOV that the sky occupies, and does it have a tool to smooth the banding in the sky? If so, any pointers to it's use? Again, tracking is nice, but if you have to tweak the mask ever few frames then a 5 minute video could take weeks to edit and a longer video even longer.


Brian

 

[yannick_c]

Never had to do that, but there are a lot of tools for masking, keeing, ..., then when the sky is well selected, apply some blur and/or noise, that could help. For sure if you make one sequence that is 5 minutes long (for tests ok, but for a real video, it will be boring), it will be a pain in the ***, but if you do a normal video with 10 sec max cuts, I'm not sure all your plan will have banding (and maybe none), and if some have, you could try to correct them. But best way is to have the right exposure at the start without banding.

 

[Raptorman0909]

I just message chinfat (Channing Lowe), who's a YouTuber that does training videos for PP etc and asked him if he'd look into doing a video showing how to do this. I'm not sure if he'll comply but we'll see. Maybe others can comment on my comment to give him some extra incentive. His link is...

chinfat


Brian

 

[nomnomnom]

Hi, is this thread dead? I read all the posts here but there doesn't seem to be a conclusion with the latest firmware. It seems like DLog is superior now because of the "fixed" algorithm that allows it to have more dynamic range?

I'm struggling with this exact problem. I find that when I shoot in dcinelike, in a typical sunset situation, my shadows are VERY noisy. Was hoping to find a solution. Here's a sample video showing ungraded and graded. If I expose any higher, the sky is going to be killed

Dropbox - Dynamic range test.m4v

 

[Tomas Wangen]

Hi, is this thread dead? I read all the posts here but there doesn't seem to be a conclusion with the latest firmware. It seems like DLog is superior now because of the "fixed" algorithm that allows it to have more dynamic range?

I'm struggling with this exact problem. I find that when I shoot in dcinelike, in a typical sunset situation, my shadows are VERY noisy. Was hoping to find a solution. Here's a sample video showing ungraded and graded. If I expose any higher, the sky is going to be killed

Dropbox - Dynamic range test.m4v

Dcinelike at ISO 100 with noise reduction will be a safe choice in most situations. The main issue with dlog 8 bit is very reduced number of tones in highlights with high potential for severe banding in uniform bright areas (such as a clear blue or white sky).

You can get away with good results with dlog if you do not have the uniform highlights, so it depends on the scene and the light.

 

[nomnomnom]

Dcinelike at ISO 100 with noise reduction will be a safe choice in most situations. The main issue with dlog 8 bit is very reduced number of tones in highlights with high potential for severe banding in uniform bright areas (such as a clear blue or white sky).

You can get away with good results with dlog if you do not have the uniform highlights, so it depends on the scene and the light.

I shot with decinelike at ISO100 but if you look at my sample video, the dark areas (trees etc) is really noisy. I don't see how I can shoot differently? As overexposing would wipe the sky

 

[Tomas Wangen]

I shot with decinelike at ISO100 but if you look at my sample video, the dark areas (trees etc) is really noisy. I don't see how I can shoot differently? As overexposing would wipe the sky

Try noise reduction during post processing. I use temporal combined with spatial noise reduction whenever I need to lift shadows.

 

[nomnomnom]

Try noise reduction during post processing. I use temporal combined with spatial noise reduction whenever I need to lift shadows.

I shouldn't have to use noise reduction at ISO100, come on now...

 

[Dronefriend]

I shot with decinelike at ISO100 but if you look at my sample video, the dark areas (trees etc) is really noisy. I don't see how I can shoot differently? As overexposing would wipe the sky

This is because D-Cinelike takes Camera Input -> Applies Gain.

What might seem to be correct exposure is in fact underexposed, and comes out noisy.

And since the Gain is applied to all the spectrum, you cannot expose higher because, as you have noticed yourself, it blows highlights.

Solution: Try Truecolor instead. Only minimal Gain applied, and only to Shadows. Lets you expose correctly IMO.

Good luck

 

[nomnomnom]

This is because D-Cinelike takes Camera Input -> Applies Gain.

What might seem to be correct exposure is in fact underexposed, and comes out noisy.

And since the Gain is applied to all the spectrum, you cannot expose higher because, as you have noticed yourself, it blows highlights.

Solution: Try Truecolor instead. Only minimal Gain applied, and only to Shadows.

Good luck

Interesting, so why do people recommend dcinelike for high contrast scene (like mine)? I'm just better off using truecolor??

 

[Dronefriend]

Interesting, so why do people recommend dcinelike for high contrast scene (like mine)? I'm just better off using truecolor??

Well, I don't know. People have different opinions. I used to use Cinelike myself. It is one of the oldest Phantom profiles and has always been better than Normal and the quirky D-Log.

Truecolor was introduced relatively recently. The only thing it does, from what I have seen, is add a small bump around the 6 % lowest tones, blending it in with the 6-12% lowest.

I find it superior because it gives a more real exposure in the viewfinder/iPhone. In Cinelike something that looks good at 100 ISO should really be 200 ISO. This is what you will notice in Truecolor. It is better to record in higher ISO than to push in post (or directly as Cinelike does).

Up to 800 ISO which is maximum if you want it to look good.

 

[skyeboysteve]

This is because D-Cinelike takes Camera Input -> Applies Gain.

What might seem to be correct exposure is in fact underexposed, and comes out noisy.

And since the Gain is applied to all the spectrum, you cannot expose higher because, as you have noticed yourself, it blows highlights.

Solution: Try Truecolor instead. Only minimal Gain applied, and only to Shadows. Lets you expose correctly IMO.

Good luck

What sharpness, saturation, and contrast settings do you usually use with Truecolor? Unfortunately, I'm no expert in experimenting myself as I really don't know what to look for; other than the fact I tend to like to saturate things more than the real pros; and I know what noise looks like.

 

[Dronefriend]

What sharpness, saturation, and contrast settings do you usually use with Truecolor? Unfortunately, I'm no expert in experimenting myself as I really don't know what to look for; other than the fact I tend to like to saturate things more than the real pros; and I know what noise looks like.

I use -1/0/0.

Limiting contrast with -2 for example will not result in better results when grading.

Let me illustrate. The codec, is a spectrum of data available for recording. I am going to call it a doorway.

Your video is a table. It is painted in the colors of the rainbow. ALL colors existing.

When you record, the table is pushed through the doorway. When using value 0, it fits as perfectly as it can through.

When using minus levels of contrast, you are compressing the table. Many fibres are lost, they fall to the ground. Gone forever, you compressed it!
It comes out on the other side. Using Post options, you can stretch it back out. But some gradients are forever lost, you squeezed them off!

If using a positive (+2) contrast value, the table will expand before passing. It wont fit. The edges will shatter/disappear. The table comes out on the other side, same size as it was before expanding. But the edges are gone. You have lost some gradients in the process.

----
The end.

It is my belief the whole -3 bonanza started because people thought footage looking gray, like Alexa footage (which is a whole other story), would automatically make it more like Alexa quality. This is rubbish. Do not compare the Phantom cam with a 70.000 dollar cinema camera.

This in combination with the former upper 60 mbit codec limit. The absolute lowest parts of the spectrum suffer in this low bitrate. It might have been a good idea to compress the table then. Lost some midtone "fibres" to save some shadows. But that does not apply anymore now that we have 100mbit recording.

Pretty satisfied with that analogy there. Hope you understand.

 

[nomnomnom]

I use -1/0/0.

Limiting contrast with -2 for example will not result in better results when grading.

Let me illustrate. The codec, is a spectrum of data available for recording. I am going to call it a doorway.

Your video is a table. It is painted in the colors of the rainbow. ALL colors existing.

When you record, the table is pushed through the doorway. When using value 0, it fits as perfectly as it can through.

When using minus levels of contrast, you are compressing the table. Many fibres are lost, they fall to the ground. Gone forever, you compressed it!
It comes out on the other side. Using Post options, you can stretch it back out. But some gradients are forever lost, you squeezed them off!

If using a positive (+2) contrast value, the table will expand before passing. It wont fit. The edges will shatter/disappear. The table comes out on the other side, same size as it was before expanding. But the edges are gone. You have lost some gradients in the process.

----
The end.

It is my belief the whole -3 bonanza started because people thought footage looking gray, like Alexa footage (which is a whole other story), would automatically make it more like Alexa quality. This is rubbish. Do not compare the Phantom cam with a 70.000 dollar cinema camera.

This in combination with the former upper 60 mbit codec limit. The absolute lowest parts of the spectrum suffer in this low bitrate. It might have been a good idea to compress the table then. Lost some midtone "fibres" to save some shadows. But that does not apply anymore now that we have 100mbit recording.

Pretty satisfied with that analogy there. Hope you understand.

That's not my what I find in my tests. I just went and shot some indoor tests with 0 all the way down to -3. I find that with -3 you get a bit more details in the highlights. The shadows are about the same. So I think it's worth shooting -3, I don't see anything lost with -3, so why not take a chance.

 

[skyeboysteve]

That's not my what I find in my tests. I just went and shot some indoor tests with 0 all the way down to -3. I find that with -3 you get a bit more details in the highlights. The shadows are about the same. So I think it's worth shooting -3, I don't see anything lost with -3, so why not take a chance.

-3 contrast?

 

[nomnomnom]

-3 contrast?

yes -3 contrast, DLog, +2EV. Under those settings, the highlights are slightly better when the contrast is added back in post.

 

[Dronefriend]

yes -3 contrast, DLog, +2EV. Under those settings, the highlights are slightly better when the contrast is added back in post.

Well yes, but if you cannot notice you have a ton of less info in everything else, you might not have a great monitor.

The new D-Log is great at preserving Highlights, but it comes at such a cost of gradient information in all other departments that it is not worth it imo.

You don't need -3 with D-Log... It would probably look like you have 4 tones in each color.

Please provide screenshots of finished graded footage.