Slightly different colors from original media to imported media

ScrapyardFilms wrote on 5/18/2021, 2:41 PM

Hello! I'm helping a user out on the VEGAS Pro Reddit forums (https://www.reddit.com/r/VegasPro/comments/nf0qhp/nikon_footage_looks_washed_out/) and I've come across a users issue that I am not sure why it's happening to them.

They have some footage recorded on a Nikon D5300 and D7000 and when viewing it in VLC, the colors look natural and as they should. When they import it into VEGAS, the colors change a little bit and lean a little on the warmer side.

What's interesting is that when the footage is imported in to Resolve, it looks identical to the colors that VLC plays leading me to believe there's some setting in VEGAS that could possibly be changed or VEGAS is interpreting this footage slightly incorrectly.

You can see how the colors are a bit different and that's clearly showed by the vectors scope. If anyone has any suggestions as to why this would happen, I'd love to hear your thoughts.

 

 

Comparison Screenshots:

 

Here's a link to download the test video:

https://wetransfer.com/downloads/5cd4dd4c32ad53a86c79f0698ec705f720210518183747/21aa4b

 

Here's a copy of the media Info of the test video:

General
Complete name                            :
Format                                   : MPEG-4
Format profile                           : QuickTime
Codec ID                                 : qt   2007.09 (qt  /niko)
File size                                : 41.0 MiB
Duration                                 : 20 s 896 ms
Overall bit rate                         : 16.5 Mb/s
Encoded date                             : UTC 2040-02-06 06:28:15
Tagged date                              : UTC 2040-02-06 06:28:15
NCDT                                     : NCTG

Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4.1
Format settings                          : CABAC / 2 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 2 frames
Format settings, GOP                     : M=3, N=12
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Duration                                 : 20 s 896 ms
Bit rate                                 : 14.8 Mb/s
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 23.976 (24000/1001) FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.299
Stream size                              : 37.0 MiB (90%)
Language                                 : English
Encoded date                             : UTC 2040-02-06 06:28:15
Tagged date                              : UTC 2040-02-06 06:28:15
Color range                              : Full
Color primaries                          : BT.709
transfer_characteristics_Original        : BT.470 System M
Matrix coefficients                      : BT.601
Codec configuration box                  : avcC

Audio
ID                                       : 2
Format                                   : PCM
Format settings                          : Little / Signed
Codec ID                                 : sowt
Duration                                 : 20 s 835 ms
Bit rate mode                            : Constant
Bit rate                                 : 1 536 kb/s
Channel(s)                               : 2 channels
Channel layout                           : L R
Sampling rate                            : 48.0 kHz
Bit depth                                : 16 bits
Stream size                              : 3.81 MiB (9%)
Language                                 : English
Encoded date                             : UTC 2040-02-06 06:28:15
Tagged date                              : UTC 2040-02-06 06:28:15

Comments

Marco. wrote on 5/18/2021, 3:06 PM

To me it looks like as if Vegas Pro reads the video correctly because there is no clipping whereas the VLC version is clipped on black and white levels.
It seems like VLC does not recognize the full range metadata.

 

ScrapyardFilms wrote on 5/18/2021, 3:08 PM

Sounds possible. Adding Levels changes and slightly corrects the values but the main concern is the color shift and why it looks identical in VLC and Resolve but not VEGAS.

Marco. wrote on 5/18/2021, 3:47 PM

I'd say that's a colorspace mismatch (in addition to the different level handling). But hard to say which one is right and which one is false (at least the Vegas Pro levels are correct).

Marco. wrote on 5/18/2021, 4:28 PM

This is what it looks like if in Vegas Pro you apply both a level conversion and color space conversion from rec. 709 to rec. 601.

This is close to the scope of your VLC version above.

So my guess is VLC and Resolve read both the color range and the color space the wrong way (seems like they treat it as rec. 601 with limited range).

wwaag wrote on 5/18/2021, 9:12 PM

I think Marco is right in that Vegas shows the colors correctly and that both VLC and Resolve treat the Nikon footage as if it were rec601 which used to be commonplace. Here are the results of applying a 601 to 709 conversion using the Channel Blend Fx. First the original.

And here's after applying the Fx.

AKA the HappyOtter at https://tools4vegas.com/. System 1: Intel i7-8700k with HD 630 graphics plus an Nvidia RTX4070 graphics card. System 2: Intel i7-3770k with HD 4000 graphics plus an AMD RX550 graphics card. System 3: Laptop. Dell Inspiron Plus 16. Intel i7-11800H, Intel Graphics. Current cameras include Panasonic FZ2500, GoPro Hero11 and Hero8 Black plus a myriad of smartPhone, pocket cameras, video cameras and film cameras going back to the original Nikon S.

sss-v wrote on 5/18/2021, 9:33 PM

Hey guys,

These are my footage, and thanks for looking into it. I would like to add that the shirt in real life is deep red, and the camera footage played directly in the camera (or on YouTube) is also deep red. So I prefer the way the footage looks like in VLC/Resolve/Nikon/YouTube and that's why I'm assuming Vegas is the one interpreting the footage incorrectly.

sss-v wrote on 5/18/2021, 9:37 PM

I think Marco is right in that Vegas shows the colors correctly and that both VLC and Resolve treat the Nikon footage as if it were rec601 which used to be commonplace. Here are the results of applying a 601 to 709 conversion using the Channel Blend Fx. First the original.

And here's after applying the Fx.


How do I apply a 601 to 709 conversion? Could you share your preset?

RogerS wrote on 5/18/2021, 9:41 PM

I'm a little confused by this part

"Color range                              : Full
Color primaries                          : BT.709
transfer_characteristics_Original        : BT.470 System M
Matrix coefficients                      : BT.601"

So it's full range with 709 primaries and BT 601 matrix coefficients? I didn't realize these could be mixed and matched.

So is the issue that the file is actually 601 but Vegas is interpreting as 709?

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

wwaag wrote on 5/18/2021, 9:58 PM

@sss-v

Here's a link to the two presets--601 to 709 and also 709-601.

https://www.dropbox.com/sh/w9sae645mxt5ihr/AABEGg5EtGQjBXTri5h-LCH9a?dl=0

 

AKA the HappyOtter at https://tools4vegas.com/. System 1: Intel i7-8700k with HD 630 graphics plus an Nvidia RTX4070 graphics card. System 2: Intel i7-3770k with HD 4000 graphics plus an AMD RX550 graphics card. System 3: Laptop. Dell Inspiron Plus 16. Intel i7-11800H, Intel Graphics. Current cameras include Panasonic FZ2500, GoPro Hero11 and Hero8 Black plus a myriad of smartPhone, pocket cameras, video cameras and film cameras going back to the original Nikon S.

sss-v wrote on 5/18/2021, 10:05 PM

@sss-v

Here's a link to the two presets--601 to 709 and also 709-601.

https://www.dropbox.com/sh/w9sae645mxt5ihr/AABEGg5EtGQjBXTri5h-LCH9a?dl=0

 

Thanks. I'm an audio person and kind of new to video and not used to working with compressed footage (audio is always uncompressed). Is applying this conversion going to degrade the footage? My alternative would be doing the editing in Resolve and no conversion is needed there.

RogerS wrote on 5/18/2021, 11:22 PM

Resolve may be doing a similar conversion internally. You'll also have a full to video levels conversion upon output so it will be changed regardless (Resolve has to do this as well).

Most video is heavily compressed and people still make adjustments to it that are far beyond just these tweaks to each channel (correcting white balance, applying creative look LUTs, conversion to B&W, etc.). Then after you render and bake in any changes the file will be recompressed and significantly degraded if you choose to upload it to a site like YouTube. That last step is where you'll really see the changes, not in minor color adjustments.

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

sss-v wrote on 5/18/2021, 11:46 PM

Resolve may be doing a similar conversion internally. You'll also have a full to video levels conversion upon output so it will be changed regardless (Resolve has to do this as well).

Most video is heavily compressed and people still make adjustments to it that are far beyond just these tweaks to each channel (correcting white balance, applying creative look LUTs, conversion to B&W, etc.). Then after you render and bake in any changes the file will be recompressed and significantly degraded if you choose to upload it to a site like YouTube. That last step is where you'll really see the changes, not in minor color adjustments.

That's why I'm trying to avoid any unnecessary changes on my side, since I'll know YouTube is going to have it's way with it! I'll be uploading uncompressed, so I guess that'll help a little.

sss-v wrote on 5/18/2021, 11:48 PM

@sss-v

Here's a link to the two presets--601 to 709 and also 709-601.

https://www.dropbox.com/sh/w9sae645mxt5ihr/AABEGg5EtGQjBXTri5h-LCH9a?dl=0

 


I tried the preset, and made the level adjustment too. But the footage, although a lot closer, doesn't look identical to VLC/Resolve so something's still missing I guess. I don't have the technical ability to match it. Any hints?

RogerS wrote on 5/19/2021, 12:10 AM

That's why I'm trying to avoid any unnecessary changes on my side, since I'll know YouTube is going to have it's way with it! I'll be uploading uncompressed, so I guess that'll help a little.

That's really a non-sequitur. Just make the changes you need.

I apply technical correction LUTs and do further corrections to every video I make. Straight out of camera colors are never acceptable for my purposes. I upload as ProRes to YouTube when it's not too long, and otherwise a decent quality mp4 done through Voukoder in Vegas. Compared to what YouTube does to it the difference between these files is minimal.

In Resolve did you click on the media, go to clip attributes, and override levels to full range? (I didn't test the video myself but that's what Marco suggested).

See if that makes it look the same.

For a player VLC is okay but I moved to MPC (Media Player Classic) Black Edition and highly recommend it. So far it's been accurate with everything I've imported.

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

Musicvid wrote on 5/19/2021, 12:12 AM

@Marco.is correct.

RogerS wrote on 5/19/2021, 1:19 AM

This is a strange one.

With levels set to full in Resolve (it properly reads the file as full range), here's what it looks like. It appears to match Windows Media Player. MPC-BE looks like it's using limited levels, so it's not accurate with this file.



Post 601 to 709 correction, Vegas matches the rest:

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

Marco. wrote on 5/19/2021, 2:03 AM

@wwaag Are you sure the Channel Blend presets are named correctly? If I compare these presets with my 601/709 LUTs they work vice versa.

Edit:
It really seems the LUT I used is named incorrectly and the Channel Blend presets are correct. Then my conclusion about the handling of the color space would be vice versa, too. Sorry for this confusion.
And if I got it right it would then confirm Roger's result.

Marco. wrote on 5/19/2021, 2:49 AM

Just having taken a closer look onto the isolated red channel only and how it behaves when color space conversions are applied.

I now really tend to think Vegas Pro is doing it the wrong way (while it treats the levels right). Without a color space conversion from 601 to 709 there is some red information clipped which will be reconstructed by this color space conversion.

So it looks like in Resolve there would be the need to get the levels corrected. And in Vegas Pro there would be the need to get the color space corrected.

RogerS wrote on 5/19/2021, 2:57 AM

Thanks Marco for your analysis.

I believe Resolve levels are fine- the waveform looks the same as VP 18. Manually overriding auto to full did nothing so it's already at full range.

(Quick verification looking at the screenshot shows the white target and speaker to have the same RGB values)

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

Marco. wrote on 5/19/2021, 3:17 AM

O.k. Then probably only VLC is using the wrong level data.

Btw - setting both the transfer characteristics and the matrix coefficients to 709 makes no difference in Vegas Pro.

sss-v wrote on 5/19/2021, 4:22 AM

So I realized if I render AVI using Sony YUV "without" applying the 601 -> 709 conversion, the colors come out just right on their own (levels seem a bit crushed though).

And, if I apply the 601 -> 709 conversion and render using AVI Uncompressed the colors still come out wrong, even though they look corrected in Vegas.

Feeling a bit frustrated. Maybe I would save time switching to Resolve and learning that rather that trying to debug my workflow in Vegas.

Any thoughts?

Marco. wrote on 5/19/2021, 5:26 AM

@sss-v 
Does this happen to you only for the clips from that one particular camera?

"So I realized if I render AVI using Sony YUV "without" applying the 601 -> 709 conversion, the colors come out just right on their own (levels seem a bit crushed though)."

Do you mean you put your unmodified source clip into Vegas Pro with no FX applied and rendered to Sony YUV from Vegas Pro?

RogerS wrote on 5/19/2021, 6:22 AM

I tried that but it looks identical to the original file. Still orange. Levels were the same when I brought into Vegas and put it a track under the original and toggled between them.

I think the answer is to add that conversion at the track level to any clips shot with that camera and you're done.

This shouldn't be an issue for DSLRs newer than that first generation a decade ago as they are straight Rec 709.

Custom PC (2022) Intel i5-13600K with UHD 770 iGPU with latest driver, MSI z690 Tomahawk motherboard, 64GB Corsair DDR5 5200 ram, NVIDIA 2080 Super (8GB) with latest studio driver, 2TB Hynix P41 SSD, Windows 11 Pro 64 bit

Dell XPS 15 laptop (2017) 32GB ram, NVIDIA 1050 (4GB) with latest studio driver, Intel i7-7700HQ with Intel 630 iGPU (latest available driver), dual internal SSD (1TB; 1TB), Windows 10 64 bit

VEGAS Pro 19.651
VEGAS Pro 20.411
VEGAS Pro 21.208

Try the
VEGAS 4K "sample project" benchmark (works with VP 16+): https://forms.gle/ypyrrbUghEiaf2aC7
VEGAS Pro 20 "Ad" benchmark (works with VP 20+): https://forms.gle/eErJTR87K2bbJc4Q7

sss-v wrote on 5/19/2021, 9:59 AM

@sss-v 
Does this happen to you only for the clips from that one particular camera?

"So I realized if I render AVI using Sony YUV "without" applying the 601 -> 709 conversion, the colors come out just right on their own (levels seem a bit crushed though)."

Do you mean you put your unmodified source clip into Vegas Pro with no FX applied and rendered to Sony YUV from Vegas Pro?

I tried that but it looks identical to the original file. Still orange. Levels were the same when I brought into Vegas and put it a track under the original and toggled between them.

I think the answer is to add that conversion at the track level to any clips shot with that camera and you're done.

This shouldn't be an issue for DSLRs newer than that first generation a decade ago as they are straight Rec 709.


I double checked. Just a note: I set Vegas to 32-bit floating point (video levels) since in certain cases the scope looked closer to the VLC footage after applying the effects, than when Vegas was set to 8-bit full range.
------------------------------------
The files:

Camera's original
Rendered AVI YUV with No Effects
Rendered AVI Uncompressed with Channel Blend and Levels

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

The results:

With VLC: The YUV and Uncompressed files look nearly identical to the original files

With MPC-BE: Uncompressed and original files look nearly identical, but YUV looks washed out.

On YouTube: YUV and Uncompressed are washed out, but the original file looks fine.

:-/