DV AVI levels to YouTube levels

[memrah]

Greetings.

I have a quick question.

Is there a program/tool that will simply take a DV AVI file captured using a Digital8 camcorder and adjust its levels (gamma, contrast, saturation, etc.) to what is suitable for online sharing, such as uploading to YouTube?

The DV files are all color corrected and look good when viewed on a TV, but as expected, they look dark and a bit under-saturated when viewed on my computer monitor.

Thanks a lot in advance for any replies.

[smrpix]

Hundreds. What are you using currently?

If the DV has been handled properly, it's in REC 601 color space which is equivalent to 16-235 in RGB. You want to expand it to 0-255 RGB. Blacker blacks inherently make the color appear more saturated. Gamma should not be a technical issue but may be an aesthetic one.

[memrah]

Hundreds. What are you using currently?

Nothing

I didn't want to start converting before figuring out which program would be the best and easiest to use. The DV files are on external hard disks and all I have on this computer is Adobe Premiere Pro, AviSynth and VirtualDub but I wouldn't mind purchasing or downloading something else.

[smrpix]

The levels filter on any of the three will do what you want. Premiere Pro has the advantage of having actual scopes (though you can use Avisynth's histogram function to check your work.) The Avisynth documentation has useful examples.

Since you did not do the initial color correcting yourself (apparently,) you will want to check your source for legal levels so you don't discard any information. DV cameras often use a wider than legal range range for recording.

[manono]

For YouTube uploads after making something for TV viewing, I use AviSynth's YLevels filter to boost the luma:

http://avisynth.nl/index.php/Ylevels

Most of my stuff is black and white. so I don't have color problems. Perhaps use Tweak to boost the saturation if it seems too low. However, just boosting the luma may be enough

[memrah]

Thank you.

Is there a plugin or script that would automate the whole process;

1. Deinterlace the DV AVI
2. Convert pixel-aspect-ratio (if necessary)
3. Readjust levels, gamma, etc.
4. Output file suitable for viewing on computer displays

(The DV files are all in PAL, by the way)

[sanlyn]

Thank you.

Is there a plugin or script that would automate the whole process;

1. Deinterlace the DV AVI
2. Convert pixel-aspect-ratio (if necessary)
3. Readjust levels, gamma, etc.
4. Output file suitable for viewing on computer displays

What does "automate" mean? Offhand, the simple answer is No. The more practical answer is that we can't answer without a sample of the video you're working with.

[memrah]

What does "automate" mean?

There are almost 100 files.
By automate, I mean batch converting these files.
Opening each file, applying the 4 steps I listed above and moving on to the next file, automatically.
Since the DV files are already individually color corrected, I am guessing (and hoping) that they wouldn't need individual attention.
All I need is go from "formatted-for-TV" to "formatted-for-computer."

[smrpix]

The two issues I see with batch encoding are 1) correctly determining aspect ratio -- are the flags correct or even present? and 2) determining whether the originals have indeed been "properly" color corrected.

[sanlyn]

What does "automate" mean?

There are almost 100 files.
By automate, I mean batch converting these files.
Opening each file, applying the 4 steps I listed above and moving on to the next file, automatically.
Since the DV files are already individually color corrected, I am guessing (and hoping) that they wouldn't need individual attention.
All I need is go from "formatted-for-TV" to "formatted-for-computer."

We still need a sample.

[Cornucopia]

Automating/batch only works if the SAME processing is done to all elements, or if the software is smart enough to guess at what OUGHT to be done. Scratch that last item: there is currently NO available software that smart. So, are ALL your elements of the same AR, with the same color balance, with the same light levels/dynamic range? If there is any amount of variability, your ideas of automation just went out the window. Even professionals have difficulty with that level of consistency.
Especially color correction. That's exactly why pros have dedicated color-correcting stations - because footage from different sources look too different to match up. And those stations require live Colorists to man them, not automated algorithms.

Scott

[memrah]

The two issues I see with batch encoding are 1) correctly determining aspect ratio -- are the flags correct or even present? and 2) determining whether the originals have indeed been "properly" color corrected.

So, are ALL your elements of the same AR, with the same color balance, with the same light levels/dynamic range? If there is any amount of variability, your ideas of automation just went out the window.

Well, every single one of them was recorded using the same camera, which was also used as the playback device during digitization. It is all PAL, 4:3. All color correction on the DV files was then done on the same system, by the same person. But of course, I guess that still doesn't guarantee 100% consistency among them all.

We still need a sample.

I'll try to upload a sample later, when I get back home.

[sanlyn]

And if some of your sources are interlaced, some progressive, some telecined, some field-blended, and so on, and so on, and so on, etc., etc., then even if you had some auto silver bullet at hand, proceed at risk. If you're clever enough to get the $$ to buy something like Adobe Premiere Pro, you probably know that Adobe can do all that stuff (although many other tools are better at some of it) except for certain aspects of total automation, and Adobe's built-in help has full documentation on various levels and color standards, including video for broadcast, standard disc, web, and PC-only. To ask a forum for specific instructions for something no one has seen (and with no detail specs posted anywhere) is, um, expecting a bit much, don't you think?

[2Bdecided]

Hundreds. What are you using currently?

If the DV has been handled properly, it's in REC 601 color space which is equivalent to 16-235 in RGB. You want to expand it to 0-255 RGB.

No you don't. You want to work in YUV throughout and stick with 16-235. YouTube's own player (like every other player) will expand this to 0-255 RGB during decoding+display.

It's worth making sure that whatever workflow you're using is treating levels correctly. Grab a DV-AVI test card and run that through everything you're doing and see if 16 ends up as pure black and 235 ends up as pure white. Sony Vegas is notorious for messing levels up.


The problem getting things to look bright enough on Youtube is that most PC monitors are set for a much lower peak white brightness than a TV screen, and when watching the video in the page (rather than full screen), it's surrounded by peak white which makes everything else look comparatively dark. On TV, you'd generally avoid peak white (unless part of the video is intentionally "blown out"), and only get close to it for the brightest objects. On YouTube, you have to get closer to peak white a lot of the time, allowing a little more clipping, and adjusting the gamma to brighten everything up a little. You still need black blacks otherwise everything looks washed out, but midtones/greys need to be brighter, and whites needs to hit peak white, not just come near to it.





Ideally you do this individually for each video, but if you have 100, drag a still from each into a graphic program and composite them into one image. Now figure out the adjustment that gets them all about right, then apply this to all 100 of the videos. Sounds easy, but subjectively looking at stills may change your opinion vs looking at videos, you should be aware of the 16-235 vs 0-255 issue when dragging the still into your graphics program and using the results in your video editor.


Cheers,
David.

[sanlyn]

Some good points, thanks for that post. I'd add that the use of histograms is tremendously helpful; I don't know why people struggle with eyeballing it. It's more effecti9ve to use monitors that are at least nominally calibrated. True, most viewers don't know what calibration is and the average user's displays are so far off the mark that I wonder why they tolerate them. Adjusting for a standard means that your work looks the way it's supposed to when properly displayed. If other users want to use "volcano mode" at all times that's their problem, not yours.

It might be easier to arrange sample frames from perhaps 20 or so of the "worst" off-kilter examples and treat them separately, rather than try to make all 100 conform to a few mavericks. That might involve making two batches of say, 20 of the bad guys and 80 of the better ones, but why should that be so terrible?

[smrpix]

@2Bdecided, I'm seeing a difference in material encoded 16-235 and 0-255 as played back by YouTube, identical encoding in TVMW5. The image below is composited screengrabs. Honest question, is there something I'm not taking into account?

[poisondeathray]

Be specific in how your test is configured. Did you mean YUV values or RGB values ? Encoded and uploaded as YUV or RGB ?

The 16-235 is expressed in in Y' .

Rec601 means Y' 16-235 , CbCr 16-240 gets converted to RGB 0,0,0 - 255,255,255 (or the reverse RGB=> YUV , Rec601 is just set of equations governing YUV<=> RGB conversions)

Vegas doesn't use Rec601 when using Studio RGB setup (it uses studio RGB matrix, which means Y' 0-255, CbCr 0-255 gets converted to RGB 0,0,0 - 255,255,255)


For the OP - 99.999999999% of the time, it's a configuration issue. On a properly setup TV and Computer/GFX card/Monitor setup they should look close to the same on a TV or Monitor . I'm guessing it's your computer setup , either your GFX card settings or monitor

[2Bdecided]

@2Bdecided, I'm seeing a difference in material encoded 16-235 and 0-255 as played back by YouTube...

If the video has different levels you will see a difference. Using the wrong levels will create a visible difference. It doesn't make it right.

I encoded a suitable testcard using x264 and uploaded it to YouTube...
http://youtu.be/r9pLqea8iDQ

As you can see (unless YouTube looks different on your PC to mine - always possible!) 16 comes out as black and 235 comes out as white*. The files (DV-AVI original and x264 encoded mp4) have whiter-than-whites and blacker-than-blacks (as labelled: 255,251,239 / 12,4,0), but the YouTube player clips them all to 100% white and 100% black, as expected.

Cheers,
David.

P.S. * - for whatever reason, 235 is mapped to 254 when played back here. Strange, but not worth worrying about.

EDIT: I uploaded the original DV too...
http://youtu.be/fvtog0kilNY
...which YouTube doesn't really like(!) - but you can see that 16 is black and something like 237 is white.


If you uploaded a video with black Y=0 and white Y=255, it would get clipped horribly at both ends of the luma range: blown-out whites and crushed blacks.

[sanlyn]

For the OP - 99.999999999% of the time, it's a configuration issue. On a properly setup TV and Computer/GFX card/Monitor setup they should look close to the same on a TV or Monitor . I'm guessing it's your computer setup , either your GFX card settings or monitor

I can agree with that 500%. Improper video levels in the source aside, most users assume that because their monitor/TV has 'digital" somewhere on the box, the product performs according to some common standard. Nothing could be farther from the truth. Calibration is not that difficult, even if you do have to get into a little research on what it entails. There are free methods that can be used (and they are pretty clunky, but they get you into the neighborhood of some sort of useable standard), and paid stuff is obviously the better way -- and easier. And as noted, if your gear is properly adjusted you are not to blame if another viewer is running their stuff at oddball factory settings. Quote/unquote: it will be their problem, not yours.

Thanks for the demo.