HD Guide for Vimeo, YouTube and the Web

[Jeff Harper]

I've used both Sony avc and Mainconcept avc for the web, and both come out perfectly fine for me. My conversions of projects from 720 60p and 1080 24p come out beautifully for DVD also.

I currently use Sony avc and set the bit rate depending on the amount of movement in the video. I follow the guidelines for Vimeo, it's really simple.

The biggest factor is the quality of my original footage. It it's good, then the web version comes out fine.

I was struggling with this issue and asked an award winning videographer what settings they used and they said Sony AVC with a bit rate of around 12 mbps, I think, which seemed very low, but their online videos look stunning.

Improper monitor settings will create a discrepancy I think also, and this has been discussed to death also. When I calibrated my monitor it helped with this issue somewhat.

I render to Sony AVC, then play back the file before uploading, and after uploading I watch the video online, and they look close to identical to me.

[Jerry Amende]

Zhong,

Aha, your videos are a great example of exactly why you want to use the cRGB->sRGB Levels FX to h.264 renders while leaving the WMV untouched. Actually, you want to map the h.264 to the 16-235 color space - the cRGB->sRGB Levels FX is a "hammer" to do this job. A better way would be to use your Video Scopes to do this.

Here's the reason: online players such as YouTube, Vimeo, JW Player, etc. will take the 16-235 range and expand it to 0-255 (you'll even see this in WMP or VLC). If your footage exceeds the 16-235 range, the player still attempts to expand them to 0-255 and the whites are blown out & blacks are crushed. Unfortunately, the Vegas Preview does not exhibit this phenomena. A temporary fix would be to apply the cRGB->sRGB FX to the "Video Output FX" on the Preview Panel. However, make sure you disable it before you render or the render will be expanded once again.

Take a look at the attached image. I tried to capture the same frame in both your h.264 & WMV versions. Yes, the h.264 looks "more vibrant", but the whites are blown out. Note the detail in the veil. I think you will agree that detail is lost in the h.264 render.

Now, maybe the "vibrant" look is the one you wish achieve and don't care about the (minor) loss of detail. In that case, go with it. On the other hand, you could map the levels and then use the Color Curves FX to increase the contrast and get the best of both worlds.

I know this is an extremely confusing subject, and I'm trying to do my best to explain it. Hope I'm helping.

...Jerry

[Zhong Cheung]

I'm confused what you mean by mapping the levels and using Color Curves FX to get the best of both worlds?

I use Color Curves on all my clips, applying an S-curve to my Technicolor Cinestyle shots (which are all of them). Then I play with each clip individually with color corrector, brightness/contrast, etc.

What does mapping the levels mean? And using video scopes instead of the cRGB to sRGB...how?

Are you saying the Sony Vegas preview window displays 0-255? And only if I apply the cRGB to sRGB filter to the preview monitor will it show 16-235?

I'm really confused...when I'm color correcting/grading, should I be viewing my preview monitor at 0-255 or 16-235? I want what I see in Vegas to look identical as much as possible to what I see when I upload to Youtube and identical to what I see when I burn a DVD (mpeg2).

It seems the mpeg2 renders, even at high bitrates, lose the color vibrancy again...just like the WMV. How can I fix this?

My Sony AVC renders -- which I did not render out with the cRGB to sRGB filter on -- look much more similar to my Vegas preview window (without the cRGB to sRGB filter on) than the WMV renders did. However, both AVC (rendered w no levels filter) and WMV (also rendered with no levels filter) look different than what I see in my Vegas preview (w/ cRGB to sRGB filter OFF). I posted a screen shot showing all three in one picture in my earlier post I think.

The current AVC render you see has the amount of vibrancy I liked and that my client preferred. Yes, you're right that it lost detail compared to the WMV, but the WMV looked so flat and dull - people didn't like that. It looked vibrant in my Vegas preview window too (w/ no levels filter), much different than the flat WMV render.

Thanks for all your detailed explanations, Jerry! Learning a lot from you.

[Jerry Amende]

Zhong,

First, the phrase "best of both worlds" was probably the wrong phase to use. I should have said, "you can use the Color Curves FX or other techniques to increase the contrast of while retaining the 16-235 color space", sorry.

Second, your questions are good, and yes, this is a rather confusing subject. I think the best way to explain this subject might be to produce a short tutorial/screencast to explain it. After all, if a picture is worth a thousand words, a video should be worth at least 10,000 - correct? This might take me a week or so to get things together as I've got some other projects going on.

...Jerry

[Stephen Crye]

Hi;

Jerry, I too would be very interested in your tutorial. I'm shooting with an AVCHD camera that I presume produces YUV420 (4:2:0 ?) video. I edit with Sony Vegas Pro. I've only recently been getting into color correction, using the scopes, etc. I'm pretty confused about the 16...235 thing, monitor gamut, the sRGB 16-235 option on the waveform scope, etc.

I'm not just sitting here waiting for you to reveal all knowledge ... I've spent hours and hours reading on the web, but there is so much information, spread out all over the place. Much of it just raises new questions ... such as this Wikipedia article:

YUV - Wikipedia, the free encyclopedia

In particular, one thing that is bugging me ... is the AVCHD 4:2:0 video produced by my Sony NX70 limited to the 16-235 gamut? Or is that even a valid question? Like the OP, I don't know if I ought to leave my monitors and nVidia cards at 0-255 or use the 16-235 setting ...

Very confused!

Steve

[Jerry Amende]

Steven,

The more I try to explain this to people the more I confuse them - that's why a video tutorial with examples may be the best way to explain it. Please be patient with me, it will undoubtably take some time.

I have two AVCHD cameras - a Panasonic TM700 & a Canon HG21. Both shoot in the 16-255 range. So when I have a project that is intended for h.264 render, I routinely use the Sony Levels FX to set the "Output End" to 0.905 which reduces the Luminance to the 16-235 range (although this may vary if the luminance does not reach 255). I use the "Histogram Scope". If you wish to test your NX70 to see what it shoot in, here's a procedure you can use: Survey: What min/max levels does your cam shoot?

Also, here's a recent thread in the Sony Creative Software Forum that may shed additional light on this subject: Color space clarification needed!

...Jerry

btw: I'm user "amendegw" over in the SCS forum.

[Stephen Crye]

Originally Posted by Jerry Amende

Steve,

... a Panasonic TM700 & a Canon HG21. Both shoot in the 16-255 range. ..

Jerry, thanks for the nice reply and the links. I will study them over the next few days. Also looking forward to the tutorial. Having done only one video tutorial (on Vegas Time-Lapse using Huffy and VeeDub, available on my stevecrye YouTube channel) , I know they are a lot of work, so it will be greatly appreciated!

I just want to make sure that when you wrote "16-255", that was not a typo. Did you mean 16-235? 'Cause if you mean 16-255 then I am *really* confused now! ;-)

Thanks again,

Steve

[Stephen Crye]

Hi All;

From one of the posts on the 'Survey" link that Jerry provided a few posts above, this had some great info that helped clarify a lot of things. BTW by the weekend I will do the Survey test for my cams, post both here and there.

---------
Subject: RE: Survey: What min/max levels does your cam shoot?
Reply by: GlennChan
Date: 6/27/2011 8:25:04 PM

There are different issues going on.

First issue: (This may be esoteric and not really matter.)

There are three colorspaces you need to know about.

Y'CbCr... as is commonly used and described in Rec. 601 and 709. (We'll assume that 601 and 709 are the same because it doesn't matter for this discussion, even though they aren't.)

studio RGB, which is RGB with black at 16 and white at 235.
computer RGB, which is RGB with black at 0 and white at 255.

2- Y'CbCr has black at Y'=16 and white at Y'=235. Suppose you are capturing from an analog tape and the calibration is a little off or there is edge sharpening going on. The code values above 235 (Y') will record those values. Later on you could potentially bring those values back into legal range. Had the standard been white at Y'=255, then you can easily get clipping and you'd never be able to bring that information back.

3- Vegas converts Y'CbCr to studio RGB for DV and HDV (in the later versions... I think 6 and above). Sometimes/oftentimes there is useful information above 235 Y'. Mapping 235Y' to 255RGB would clip all the information above 235Y'.
Also, almost all video cameras have useful information above 235Y'.

4- However, this does not mean that all information is necessarily preserved. Yes, 255 Y' will get mapped to 255 RGB'. However, Y'CbCr color space still far exceeds computer RGB. Y'CbCr can represent a lot of crazy colors, some of which can't exist because it would require the existence of negative light. In practice, some cameras will record Y'=255 while the chroma is not neutral, so the equivalent RGB value will have one channel that is above 255 RGB.

Clipping can still occur when converting from Y'CbCr to computer RGB. I believe that some cameras actually do record colors that fall outside computer RGB (though I could be wrong here).

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

Second issue:

The values in Vegas depends on what codecs you are using and how they are configured.

For example, if you set Vegas to use the Microsoft DV codec (not recommended), DV footage will decode to computer RGB.... 0-255. And you will get clipping.

If you set Vegas to use 32-bit in project properties and use the full range setting, then HDV will decode to computer RGB. You will have illegal values above 255 RGB, though you can bring this information back since 32-bit floating point numbers are being used.

It may be more useful to say what default codec Vegas uses to decode footage from that camera.

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

Third issue:
Some/most cameras record values outside the legal range, and/or do not follow standards at all.

Almost all video cameras will record information above 235 Y'.

I can't remember if some of the dSLRs will put black level at Y'=0 and white at Y'=255.
----

Steve

[Jerry Amende]

"I just want to make sure that when you wrote "16-255", that was not a typo. Did you mean 16-235? 'Cause if you mean 16-255 then I am *really* confused now! ;-)

Yup, both my TM700 & HG21 shoot in the 16-255 color space. If you look in the Survey: Survey: What min/max levels does your cam shoot? post, that's my TM700 in the third post down and my HG21 in the 7th post down. Incidentally, I don't think you'll see the embedded Video Scope images unless you're a member of the forum, so I'll attach the images here.

For those that haven't read the "Survey" link, here's the methodology:

"Black footage method: Cover the lens with the lens cap or whatever, close the iris, set gain to zero, set the shutter as fast as possible.

White footage method: Open the iris as wide as possible, set gain to max, set the shutter as slow as possible, remove any ND filters, point camera at sky."

...Jerry

btw: I'm continuing to work on-and-off on a video presentation, but it's slow and I don't have a whole lot of time to spend on it.

[Stephen Crye]

Hi Jerry + Everyone;

The "A New Look" method et al (which I became aware of a month ago and spurred me to try to understand concepts related to the "16-235" alluded to in the supporting online material) now is finally starting to make more sense.

I just completed the white/black levels tests for my NX70. The whites are pretty much full 255, but the blacks are in the 20-ish range. I found it odd that the histogram for the white was just solid bars, but the blacks looked like little mountain peaks. I guess the NX70's factory settings don't fully crush the blacks (good thing Vegas is ready to rescue) I hovered the cursor over the middle of the groups so that the value can be read in yellow on the far right. For the FullWhite, the histogram just showed solid vertical bars, so I had to streeeeetttccchhhh the vectorscope window across two monitors to get the screen shot attached.

Tomorrow I hope to do the same test for the CX550V, which despite the lack of decent manual controls and some other bells and whistles, seems to have better image quality.

So, I can see how it would help the video to look better on YT if one first adjusts tonality so the blacks go to zero and not below, the whites hit 255 and not higher, and then apply the cRGB >> sRGB FX to "squeeze them in anticipation of YT stretching them out.

What still confuses me is the Vegas preview window display, and the little setting on the Vegas vectorscope for sRGB. Should I leave the vectorscope set for full 0-255 while I adjust tonality and color?

Finally, I do a lot of rendering for Blu-Ray. Do I need to worry about the cRGB>>sRGB for Blu-Ray before rendering? (I have never used it in the past and the BD's seem to have good tonality)

Thanks!

Steve

[Nick Hope]

Stephen, this is a useful result. I'll add it to the list on the SCS forum. (Incidentally, there is an important firmware update on the way for your cam, if you didn't know)

Looking forward to your CX550V results too.

What still confuses me is the Vegas preview window display, and the little setting on the Vegas vectorscope for sRGB. Should I leave the vectorscope set for full 0-255 while I adjust tonality and color?

I set it to "Studio RGB (16 to 235)". That's perfect because then the 0 and 255 lines then become your black/white limits for both web and TV delivery, but you can still see the illegal values.

I have an external monitor (Windows Secondary Display) set with "Use Studio RGB", so I can see what the luminance should look like when delivered to the web or TV, but at the same time I monitor the Vegas preview window which uses computer RGB, so I can see what detail might be getting clipped from the highlights and lowlights.

Perhaps the easiest way to think about this levels issue, is that Adobe Flash Player (in the majority of cases) does the same thing as a TV-delivered scenario (e.g. DVD). In my opinion that's great. The problem is that users of "0-255" cams such as Canon dSLRs, editing in Vegas, don't realise that their shadows and highlights will get clipped unless they adjust levels to prevent that.

A final thought... One or two of my DV clips lead me to believe that some of the "16-235" cams might give you blacks below 16 in some high-contrast shots. So the "black out" test technique may be flawed. I'll do some testing on that.

[Jerry Amende]

Nick,

Thanks for registering in on this discussion, 'cuz...

1) You are much more expert in this subject than I, and
2) I'm having a difficult time conveying this concept and another voice hitting it from a different angle can only help.

...Jerry

[Jerry Amende]

Stephen Cry said, 'Finally, I do a lot of rendering for Blu-Ray. Do I need to worry about the cRGB>>sRGB for Blu-Ray before rendering? (I have never used it in the past and the BD's seem to have good tonality)"

I don't do a whole lot of DVD/Blu-Ray work, but I believe the the stardard practice is to confine the color space to 16-235 (as Nick confirms in his post).

Here's what I do for projects I'm not sure about. Nick has come up with a clever test - I call it the "Nick Hope Levels Test" - which illustrates this phenomena.

Nick Hope Levels Test on Vimeo

When the left side of the video looks all white & the right side looks all black, then your color space must be confined to the 16-235 range. I put together a Vegas Project with generated media only (attached). If I want to determine how a Blu-Ray project performs, I merely render this out to Blu-Ray, insert the test into my player and see what the whites & blacks look like.

...Jerry

[Stephen Crye]

Hi Nick!

I think I remember you from the creative "Bovine" forums if not, then I have a strong false memory of chatting with you, watching some great underwater clips of yours.

Don't get me started about the NX70 firmware upgrade - far too little. In another thread I'm going to post a list of *MUST* have firmware fixes for the NX70 - annoying things like not being able to use the viewfinder while the LCD is open (which forces me to constantly open and close the LCD in bright sunlight to get anything done), and the need to have the ability to adjust the update speed of the auto-focus to prevent it from trying to focus on noise in high-gain situations.

And for Sony to think that just giving us the ability to set the zoom rocker to "slow" is a fix for the horrible problem is a joke. It is supposed to be a *variable* rocker, after all!

All we want is for this $3200.00 cam to work at least as well as the little MC50 (aka CX550V) ! grrr.

Sorry to rant.

Steve

[Stephen Crye]

Hi Everyone;

OK, I've been doing some tests regarding the purported "stretching" of levels done by YouTube when YT processes an uploaded video. My results are (not unexpectedly) confusing.

My test:

I created a short clip in Vegas Pro 10e. With the Waveform scope set to Studio RGB 16-235, I used the Sony Levels FX plugin to squeeze the whites down to 100% and lift the blacks up to 0%. I used the Output Start and Output End sliders to bring down the whites and lift the blacks. I then rendered using Sony AVC to an .mp4. I uploaded two versions of the clip, one with the Levels plugin engaged and one without.

Visually, the clip without the levels plugin had the look of crushed blacks and more "blown-out" highlights. My monitor is set to Computer RGB 0-255, and I use Calibrize to calibrate it. In the clip with the levels adjusted, the blacks were less crushed and the highlights less blown out.

After I uploaded the clips to YT, I then downloaded them and brought then into Vegas for comparison. I took a screen grab of the Waveform scope at about the same point in the clip, in the middle where a woman walks in front of the performers. Those four screen grabs are attached to this post.

All four versions of the clip are available on my website for download:

http://www.huecotanks.com/temp1/Danx...oStudioRGB.mp4
http://www.huecotanks.com/temp1/Danx...RGBAfterYT.mp4
http://www.huecotanks.com/temp1/DanxNoLevelAdjust.mp4
http://www.huecotanks.com/temp1/Danx...ustAfterYT.mp4

The YouTube videos (unlisted) are here:
DanxNoLevelAdjust.mp4 - YouTube DanxLevelsAdjustedToStudioRGB.mp4 - YouTube
My Confusion: Although YT seems to be "stretching" the levels, it is subtle, and not nearly what I expected. I expected the YT-processed "Studio RGB" version of the clip to look more like the unadjusted version of the clip. Instead, the YT processed versions look pretty much like the way they looked before YT processed them.

I welcome comments and analysis from any and all gurus ...

Thanks,

Steve

Attachments : top row = after YT , bottom row = before YT, left = no adjust, right = adjusted