Newly discovered (but old) Vegas MOV bug!

Laurence wrote on 4/26/2013, 11:16 PM

I just realized today that Sony Vegas 12 (and previous versions) are expanding the colorspace range of MOV video clips. That is that if you take a sRGB range video clip in Sony Vegas, it will display it as though it was cRGB.

I discovered this accidentally by trying to figure out why ProDADs Mercalli SAL v3 was compressing the colorspace range of my MOV clips. After about a week of back and forth with the ProDAD tech department and Facebook page, I finally figured out today that I had it completely backwards. The MOV clips color range was what was displaying incorrectly. The converted clips had fixed a problem, not created one!

All you guys (me included) who have DSLRs that shoot in MOV format and have wondered why Canon or Nikon chose to shoot their video in cRGB range...well we were wrong! These cameras have been shooting in sRGB range all along. It's just that Vegas has been expanding these levels!

Back to post

Comments

Laurence wrote on 4/27/2013, 2:25 PM

Doesn't the GH2 just have the AVCHD modes? AVCHD mode on the GH3 is fine. It just the MOV format that I have an issue with.

Marco. wrote on 4/27/2013, 2:31 PM

Yes, what I meant were the GH2's AVCHD files. But the container doesn't matter here because Vegas Pro 12 decodes both kind of video – GH2 AVCHD and GH3 MOV – using Sony AVC.

To me it really seems like the recording modes of GH2 and GH3 differ in the way GH3 uses an expanded range of values especially for the blacks whereas GH2 really tries to set reference black to 16 and reference white to 235 allowing only swings and specular highlights to expand this range.

farss wrote on 4/27/2013, 4:14 PM

Musicvid said:
"It was actually a discussion of the different MP4 decoders in Vegas 8 and 10, and may be the key to why Laurence's examples are giving different results in different Vegas versions."

Exactly what the tests results I posted yesterday reveal.

In Vegas 9 the MOV and MP4 files have the same 0-255 levels

In Vegas 10 and 11 the MOV file still reads 0-255 BUT the MP4 file reads 16-235.

In a Skype session with Grazie yesterday he confirmed V12 does the same.
Conclusion:
Between V9 and V10 the way MP4 files are decoded changed.

Laurence said:
"Here was my thinking in doing this experiment. If I am right and the camera is recording sRGB levels but Vegas is stretching them out to cRGB, then a bright window should probably go a little over the 235 mark and be clipped after the stretching. This brightness would be clipped and not recovered when you scaled the levels to sRGB range."

You are being fooled by using the Histogram, they can indicated clipping when there isn't any. Quite simply histograms are not a reliable tool for reading levels.
The only reliable tool is a waveform monitor. I see waveform monitors and vectorscopes everywhere in post houses and TV stations. Histograms and Parades are rare. They have their place but not for reading levels.

Bob.

Marco. wrote on 4/27/2013, 5:40 PM

Yes, I also think histograms should be used very, very carefully. To me a histogram is kind of forensic analyzer but it isn't a good choice to get out where reference black and reference white is set. Waveforms are the primary tools. Then histograms can unveil some fine extra details. It's like sorting out some nano particles in a painting.

NormanPCN wrote on 4/27/2013, 6:22 PM

When using the Waveform to view file input range, what should the video scopes studio RGB 16-235 option setting be? Checked or cleared?

When checked it seems to do a studio to computer rgb adjustment.

musicvid10 wrote on 4/27/2013, 8:34 PM

True, but the histogram is a great visual tool for quickly setting starting and ending luminance values. That and my eyeballs usually get the job done.
Granted, it is not a precision grading tool.

NickHope wrote on 4/28/2013, 12:51 AM

I use all 4 scopes when I'm colour correcting. My eye is constantly flitting between all 4 scopes, and I can't imagine doing without any of this info now.

An old 17" monitor is perfect for this. I have 3 tabs set up on that window so that when I'm not colour correcting that whole screen shows the Media Manager, or sometimes Edit Details.

I agree that the histogram can be misleading. On the Vegas histograms, a few pixels can give the impression of being a lot more than it really is. In Laurence's last screen shot, the stuff to the right of the peak looks to me like a bit of "spillage" after Mercalli has done its thing, as opposed to evidence of hidden detail.

(Thanks for the link musicvid. I must remember to drop the odd naughty(ish) word next time I want to find something without bookmarking it.)

deusx wrote on 4/28/2013, 3:08 AM

GH2 has settings to shoot sRGB or AdobeRGB. I assume GH3 can do the same.

This applies mostly to stills, but depending on which mode you use when shooting video it is apparently possible to shoot video using AdobeRGB.

Could this be the cause? You are accidentally shooting using AdobeRGB and that throws off something somewhere in the pipeline.

Laurence wrote on 4/28/2013, 6:21 AM

Adobe RGB and Studio RGB is just a stills parameter I believe. I don't think it has any effect on video.

deusx wrote on 4/28/2013, 6:44 AM

apparently it is possible to use AdobeRGB when shooting videos. It depends on which video mode/settings you chose. Since it is a hybrid camera settings will overlap under certain conditions.

Laurence wrote on 4/29/2013, 2:38 PM

OK, now that I know how to do RGB luminence histograms, let's see if this finally proves my Vegas remapping theory:

Native Nikon D5100 footage on a Vegas timeline:

D5100 footage after Mercalli:

Laurence wrote on 4/29/2013, 2:42 PM

What this shows is that there is red and blue information that is being chopped of when you look at a native D5100 mov clip on a Vegas timeline.

If the D5100 video was actually in the cRGB range that Vegas displays, this information wouldn't be there to begin with. I seriously doubt that Mercalli is generating this information that would have to be in the "over 255" in the red and "under 0" in the blue range that would have to be the explanation if Vegas wasn't remapping the levels!
Have I finally proved it this time guys? Mercalli isn't compressing the levels. It's fixing a Vegas mistake!

farss wrote on 4/29/2013, 4:17 PM

"Have I finally proved it this time guys?"

You have wasted a lot of time and proved nothing.

You cannot use histograms to determine clipping, period.
Historgrams display statistical distribution of values.
Seeing as how you don't believe me and others please read:

http://en.wikipedia.org/wiki/Histogram

I long ago posted the results using Vegas's waveform monitors.
They show only mild clipping in the area of the paving in full sun in the source footage.
After processing by Merc and displayed in Vegas's waveform monitor the highlights remain EXACTLY the same, there is no magic highlight detail recovery or any such thing going on. Only the level has changed.

There are only two tools that can be used to correctly check levels and to look for clipping.

1) Waveform monitor.
2) RGB Parade,

Bob.

Marco. wrote on 4/29/2013, 4:24 PM

Believe it or not – it's Mercali which is pushing down levels of some video formats like MP H.264.

farss wrote on 4/29/2013, 4:34 PM

"Believe it or not – it's Mercali which is pushing down levels of some video formats like MP H.264."

PROBABLY WRONG

The same files decode differently in Vegas 8 and Vegas 9 i.e. they decode at the same Computer RGB levels as the MOV files going into Mercalli. It is only in V10,11 & 12 that they decode differently.

Bob.

Marco. wrote on 4/29/2013, 4:52 PM

Deleted as it overlaps with another thread.

NickHope wrote on 4/29/2013, 11:36 PM

Instead of guessing at this, get a copy of musicvid's grayscale test chart, render some footage from it, and test that.

Laurence wrote on 4/29/2013, 11:41 PM

But it is specific camera formats. How would I render to say a Nikon D5100 format, or a Panasonic GH3 format, or a GoPro format?

Grazie wrote on 4/29/2013, 11:56 PM

Seen it, but where's the DB link?

Cheers

Grazie

NickHope wrote on 4/30/2013, 12:01 AM

But it is specific camera formats

Good point. But where camera format is not involved in something you are testing, it makes things more precise.

where's the DB link?

Right click > Save as!

For future generations, this discussion continues here.

Grazie wrote on 4/30/2013, 12:07 AM

Done! Thanks Nick.

G

Marco. wrote on 4/30/2013, 3:13 AM

Nick, I have a bunch of test signals like the greyscale but it's rather tricky in this case because it seems like there is only a very specific format which is causing the issue. It's not MOV in general, nor MP4, and it's not h.264 as compression format in general.

In addition Mercali definitely pushes down levels of certain formats (like a Quicktime PNG MOV file) and it really looks like as if Vegas Pro 12 expands levels of a very certain type of MOV h.264.

Trying to get this down to a point where we can say how to properly handle a certain case is tough.