HD For Indies

UPDATED SUNDAY, JUNE 25TH: scroll down to where it says "SYNOPSIS OF CURRENT STANDING:"- got some new answers...

This is the follow on to the above post, where I was trying to remove 3:2 pulldown from footage acquired without timecode (so no A frame detection in CinemaTools other than relying on timecode, which was wild in this case, and therefore useless).
Synopsis: After Effects, either versions 6.5.1 or 7.0 on Macintosh, has problems writing out 10 bit QuickTime files. Or more accurately, it can create 10 bit QuickTime files, but it doesn't contain 10 real bits of data. Whaaaa? Read below for the details, it is important.

I've been writing for a while about the possibility of using After Effects as a high quality finishing tool, but now that usefulness is diminished - After Effects can make good 8 bits/channel images, but not 10 bits/channel (update: that isn't a correct analysis - see bottom).

Continuation from other article above...

....so I needed a different tool, a tool that could find and identify 3:2 pulldown. A-ha! I have such a tool - it is Adobe After Effects. If you import a file into After Effects, highlight the file in the Project Window, and press Command-F (or right click and select "Interpret Footage=>Main"), you get a dialog that has an ever so convenient button labelled "Guess 3:2 Pulldown." And right next to it, "Guess 24pA Pulldown." Heaven! You just click it, it analyzes the footage, identifies the pulldown, and from that point forward in After Effects you can treat that 24pon60 clip as a 24p clip, since After Effects will handle the pulldown removal for you in any composition you place that clip in. This works like a charm!

Here's the "before" dialog:



Here's the "after" dialog:



"Hang on" you say, "but After Effects has no batch tools, so you'll have to process these clips one at a time! What a hassle!"

Incorrect, my friend, if you know a few tricks. Allow me to demonstrate:

1.) While it is possible to right click on a clip in AE after fixing the pulldown as described above and select "Interpret Footage=>Remember Interpretation" with the intent of using "Interpret Footage=>Apply Interpretation" that would be a mistake in this case - you'd be applying this particular pulldown pattern to all the other clips, but since they also start their cadence randomly, odds are 4/5 of them would be WRONG. So what to do:
-highlight clip
-Command-F
-click Guess 3:2 Pulldown
-press Return/Enter
-down arrow (selects next clip)
-repeat until done

2.) Now that you have all your clips all set up, you're ready to get to work. But wait a minute, you'd need to make comps for all these, right? And that's a pain to set up a new comp, etc, right? Not necessarily...do this:
-select the first clip
-drag it's icon to the Create A New Composition icon - it's the little filmstrip looking icon at the bottom of the Project window, third from the left after the search (binoculars) and New Folder buttons. It is here:



-this creates a new comp that matches the footage you dragged to it - the frame size, frame rate, aspect, duration, etc. are all set to match the clip you made the comp with. Major time saver.
-with your new comp highlighted, press Command-M to Make Movie
-this opens up the Output Movie To dialog - find or make a folder where you'll want ALL your movies to go for this entire batch we're about to create.
-now go up to Edit=>Templates=>Render Settings and create the render settings you want for your batch, and then set the Movie Default to this setting.
-similarly, go to Edit=>Templates=>Output Module and edit/create the output settings you want to use for your batch, and change the Movie Default to this setting
-you now have fixed three things: your render settings, output settings, and render locations have all been set for future rendered stuff.
-now go and delete that composition you made a minute ago. No crying, it'll come back
-NOW that you have render settings, render destination, and output settings all preset the way you want them, highlight ALL the clips you want to render out as 24p in AE and drag'em to the Create A New Composition icon
-you'll see a new dialog you haven't seen before - the New Composition From Selection dialog. Looks like this:



-click on Multiple Compositions (not Single), and also check the box for Add To Render Queue (YES you want that)
-OK, forgot to mention this step, you just need to do it before rendering: right next to the Create A New Composition button it probably says 8 bpc. Option-click on it until it says 16 bpc. This is 16 bits per channel - this means if you render to a 10 bit codec, you'll get more than 8 bits of precision. This slows down the render, but if you're working with greater than 8 bits per channel footage, or want greater than 8 bits per channel of fidelity, you need it. If your source is 8 bit footage, no need to do this step for 3:2 pulldown removal, however.
-now you'll have a ton of new comps, all set to the individual settings of each clip, and the render destination, render settings, and output settings are all correct, and all these files are already in the Render Queue. Click Render in the Render Queue and take a break, it's all cookin' down for you now.

I was all pleased with myself when The Little Voice of Doubt began to whisper in my ear -

"Yeah, but is all that 10 bit goodness still working?" I had rendered to the BlackMagic 10 bit Uncompressed (4:2:2 color sampling) codec. I wanted to be sure this was all working right - I've had issues with Compressor and 10 bit files in the past, so I wanted to be sure.

To check, I made a new compostion, 2048x1080 pixels, added a solid, cranked the project bit depth to 32 bits per channel just to be sure (option clicking on 8/16 bpc to do it), and added a Ramp. The ramp was set to go from pure black to pure white, starting at 0,0 and ending at 2048,0. Why this way? 10 bits per channel is 1024 individual gradations of color. (8 bits is 256, BTW). With a 2048 pixel wide comp (2x1024), every other pixel should increment one brightness value on a 10 bit scale. So pixel #80 from left edge of comp should have a value of 40, pixel 256 a value of 128, etc. - whatever the location on screen was from left to right, the brightness readout on a 10 bit scale should be half of that.

So I set this up, and rendered it out of After Effects 6.5.1. In order to get greater than 8 bits of detail in your rendered file, in the Output Module Settings you have to select Trillions from the Depth pop-up in the Video Output section. After Effects 6.5.1 allows this if you're rendering to a BlackMagic codec QuickTime movie, but for reasons at that time unclear, AE7 does NOT (and holding down any combo of modifier keys doesn't help).

So here's AE 6.5.1 dialog letting me use Trillions:



...and here's AE 7.0 NOT letting me use Trillions:



So I tried rendering in various formats, and then imported the results into AE 7 and zoomed way in on the file, I set the color readouts to a 10 bit scale,and watched the Info readout carefully as I panned a cursor across. And got surprising results.

If it worked right, every other pixel should jump 1 unit of brightness. Here's what I found from rendered results:

BlackMagic 10 bit (4:2:2 uncompressed Y'CbCr codec): NOT - jumps of 4, seemingly dithered
BlackMagic 10 bit RGB (4:4:4): NOT - jumps of 4, seemingly dithered
Sheer 10 bit 4:2:2: NOT - jumps of 4, seemingly dithered
Sheer 10 bit 4:4:4 YUV: NOT - jumps of 4, seemingly dithered
Sheer 10 bit RGB 4:4:4: shows all black in comp, but the QT file opens and looks OK (I've contacted the developer about this possible bug)
Cineon/DPX sequence: YES, brightness increments as expected if you set the options correctly
Photoshop file sequence: YES, brightness increments as expected, when output set to Trillions in Output Module Settings, it makes 16 bit per channel Photoshop files

OK, that's sucky and perplexing - After Effects 6.5.1, even though you can set it to render Trillions to a 10 bit codec, WILL create a file that is 10 bits, but there is NOT 10 bits of precision in that file! This means After Effects is useless to directly generate 10 bits of precision in it's work. Later I figured out AE 7.0 has the exact same problem.

Hmm....could I be wrong? I tried something else - I created a simple 100x100 pixel composition, made a 10x10 pixel solid, and sampled a color from my ramp that I knew was "between the steps" in an 8 bit scale - RGB values of 339 on a 10 bit (0-1023) scale. I rendered that out in AE 7 to the BMD 10 bit codec, and it didn't work - the brightness shifted to 333. Well, not exactly - the red, green, and blue values varied slightly, centering around 333, as if it were dithered. I did another test making additional squares with brightness values of 338, 337, 336, 335, etc. Rendering to the BMD 10 bit codec in AE 7, they all shifted, and the new values were centered around brightness jumps of 4, but the RGB values varied, leading me to think it was 8 bits of precision dithered down.

I went back and did further testing with AE 6.5.1 and AE 7.0 - neither can render a proper 10 bit QuickTime with Sheer or Blackmagic codecs. Next up: Apple & AJA codecs.

So I uninstalled (just compressed to a .tar archive and deleted the BMD codecs from the Libary:QuickTime folder) and then downloaded and installed the codecs from AJA.

If it didn't before, now whether I can select Millions of Trillions in the Output Module depends on the codec I have selected. Details on AJA codecs in After Effects 6.5.1:

AJA 10 bit log RGB codec: Millions not Trillions available, and it defaults to RGB+Alpha, so have to set it back to just RGB

AJA 10 bit RGB codec: Millions not Trillions available (this codec also has a video or full range option, I chose full range 0-1023 for this test)

AJA 2Vuy codec - defaults to Millions of colors+, no RGB only offered apparently?

AJA 2vuy codec - defaults to Millions of colors+, no RGB only offered apparently? (yes, this one's apparently different, v not V)

AJA v210 codec - defaults to Millions of colors+, no RGB only offered apparently?

Apple FCP 10 bit Uncompressed - Trillions is an option

Apple FCP 8 bit Uncompressed - tested just to see what happens, Trillions IS an option

Test Results: none of'em worked right for 10 bit output. Not in AE 6.5.1, not in AE 7.0, not with 16 or 32 bpc selected, nothin'. The only way I could get true 10 bit levels of detail out of After Effects was with Cineon or Photoshop image sequences, which are a bear to work with in Final Cut Pro, and not practical for editorial workflow.

OK, that's enough testing for now, I'll see if I can dig up any more info pertaining to WTF.

Am I doing something wrong here? If you have salient info, please let me know!

-mike

SUNDAY JUNE 25TH UPDATE:

So I started looking into what's up with the bit depth issue.

First up, AJA codec support. Reader Joe Rice was kind enough to point out this AJA support article to me, which reads in part:

Q: Does KONA 2 support 16-bit renders from Adobe After Effects?

A: Yes, KONA 2 codecs have full support for After Effects 16-bit/component ('b64a') mode. However, because of how After Effects works, it can be slightly complicated. When you export a QuickTime movie from After Effects, you have to choose a codec to use (Render Queue:Output Module:Compression Settings). To use the clip with any of the usual uncompressed 4:2:2 hardware cards out there (including KONA 2) you would typically choose either Final Cut Pro's "Uncompressed 8-bit 4:2:2" (2vuy) or "Uncompressed 10-bit 4:2:2" (v210) codecs, or the board manufacturer's proprietary uncompressed codec (e.g. "AJA Kona 2Vuy Codec"). Of all these choices, the only codec that does NOT currently support 16-bit RGB is Final Cut Pro's "Uncompressed 8-bit 4:2:2" codec. The FCP 10-bit codec DOES support 16-bit RGB, as do the KONA 2 codecs.
Note: After Effects does NOT automatically determine codec settings! You may have to edit your "Adobe After Effects x.x Prefs" file (where "x.x" is your current AE version, e.g. 6.5), located in your user/Library/Preferences folder. Use TextEdit to open the file (you might want to keep a copy of the original Prefs file just-in-case you make a mistake in editing), then search or scroll down to the section of the file labeled ["QuickTime 64-bit Output Codecs"]. To enable a given codec type (for example 10-bit uncompressed), look for a line of text like:"v210" = "1". For the AJA KONA 2Vuy codec there should be a line that looks like:"2Vuy" = "1". If the codec type is missing, just add the line to the list. If the codec type is listed, but is set to "0", change it to a "1" to select it. Save the "Adobe After Effects x.x Prefs" file, and then re-launch After Effects to use the new preferences.

You may also want to add the same lines to the After Effects Pref file in the ["QuickTime 64-bit Input Codecs"] section immediately preceding the Output Codecs. This will also allow you to import 16-bit RGB files from compliant codecs.

To allow trillions of colors with the KONA RGB codec, add the following lines to BOTH the "QuickTime 64-bit Input Codecs" and "QuickTime 64-bit Output Codecs" section of the After Effects Preference file:
"R10k" = "1"
"R10g" = "1"

For the Windows version of After Effects, the same lines need to be added to the "Adobe After Effects 6.5 Prefs.txt" file, located in your Documents and Settings//Application Data/Adobe/After Effects/Prefs directory.


So I did that.

As for the BlackMagic stuff, I did a search over on their site, and found this link in their support section, which reads in part:

Trillions of Colors in After Effects

After installing or updating After Effects, I cannot find a way to render to Trillions of Colors with DeckLink. Millions is the only option available. How do I make it work?

When you install current DeckLink drivers, it edits the After Effects preferences file to enable the use of trillions of colors when rendering with DeckLink cards. When you go to the Output Module Settings and click on the Depth popup menu, both Millions and Trillions should appear as available options.

If you only see Millions of colors, then it may be that After Effects was installed after the DeckLink drivers. Simply uninstall and reinstall the current DeckLink drivers and next time you open After Efects, you should be able to render to Trillions of colors.


....so while I was already on their site, I pulled down the latest drivers - they came out with a new rev of their Multibridge Extreme drivers on the 19th, v 5.6.1, so I grabbed that too, and installed it.

If the BlackMagic drivers are installed, they take over from the Apple Final Cut Pro Uncompressed codecs. I don't know if they are literally the same, or if one's taking over, but in any case, I can't be sure of using the Apple Uncompressed with the BMD codecs installed, so I just hide'em (compress to archive) while working with AJA codecs.

I'll do some more testing later today to see what effect this has, but I need to go run an errand right now...

OK I'M BACK:

...and here's what I've learned.

It's all about your After Effects Prefs file. As described above, it depends on whether your "Adobe After Effects 7.0 Prefs" (or whatever version you're running) is properly configured or not. How can you be sure your prefs are properly configured?

If you're using AJA codecs, you have to manually go in and configure as described above (somebody correct me if new installers take care of this for you). I gotta redo some stuff to test those and the Apple codecs, I'll post that up shortly.

If you're using BlackMagic codecs, you just need to have run the BlackMagic installers AFTER you've installed After Effects. Even if that means running the installer for the version you're already running, THAT'S OK. Part of the install script doctors your Adobe After Effects Prefs file to get it right. I rendered, I tested, and the BlackMagic 10 bit 422 and 10 bit RGB codecs DID render properly to 10 bit (10 bit image fidelity was maintained).

If you're using Sheer codecs (from BitJazz), then...it's not quite so obvious.

Even after installing the latest version, it DOES doctor your AE prefs (or is it installing the Output Modules that Sheer includes that does it?). BUT...it doesn't seem to do it right, or at least mine wasn't giving me Trillions options on any of the Sheer codecs. So I went into the Adobe After Effects Prefs file, did a search for "QuickTime 64" to find the right part of the prefs file, and saw that there was now "Shr0" through "Shr7" listed, but they all were set to "0" (meaning no 64 bit/Trillions support). I also received a prompt email from Andreas Wittenstein, the developer, who informed me that when I set the Sheer codec preferences to "Perfect" instead of "Best" conversion something in the dialog, that generated an error, since "Perfect" only worked on 16 bit codecs, not 10 bit codecs, so "Best" was the preferable option, otherwise it failed to work. After that I was able to render OK...but got 8 bit dithered results.

So I did a web search for "Shr1" and found a document on Andreas' site that gave me a clue as to what codecs were 8 bit and which were 10 bit. This page lists Shr1, Shr2, Shr3, and Shr4, and they are all definitely 8 bit formats and should NOT have Trillions available. But the others may be the 10 bit codecs, so I changed all those to "1" to enable the Trillions option. After changing that setting in prefs file and relaunching After Effects 7.0, Trillions was now an option for the Sheer codecs, but I still don't think I've quite got something configured right.

A-ha- then I found the below from this page here:

imco:Shr0:BtJz "Sheer"
imco:Shr1:BtJz "Sheer RGB[A] 8b"
imco:Shr2:BtJz "Sheer Y'CbCr[A] 8bv 4:4:4[:4]"
imco:Shr3:BtJz "Sheer Y'CbCr[A] 8bv 4:2:2[:4]"
imco:Shr4:BtJz "Sheer Y'CbCr 8bw 4:2:2"
imco:Shr5:BtJz "Sheer Y'CbCr[A] 10bv 4:4:4[:4]"
imco:Shr6:BtJz "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imco:Shr7:BtJz "Sheer RGB[A] 10b"
imdc:Shr1:BtJz "Sheer RGB[A] 8bf"
imdc:Shr7:BtJz "Sheer RGB[A] 10bf"
imtc:2vuy:Shr0 "Sheer Y'CbCr[A] 8bv 4:2:2[:4]"
imtc:2vuy:Shr3 "Sheer Y'CbCr[A] 8bv 4:2:2[:4]"
imtc:Y216:Shr0 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:Y216:Shr6 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:r408:Shr0 "Sheer Y'CbCr[A] 8bv 4:4:4[:4]"
imtc:r408:Shr2 "Sheer Y'CbCr[A] 8bv 4:4:4[:4]"
imtc:v210:Shr0 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:v210:Shr6 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:v216:Shr0 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:v216:Shr6 "Sheer Y'CbCr[A] 10bv 4:2:2[:4]"
imtc:v408:Shr0 "Sheer Y'CbCr[A] 8bv 4:4:4[:4]"
imtc:v408:Shr2 "Sheer Y'CbCr[A] 8bv 4:4:4[:4]"
imtc:v410:Shr0 "Sheer Y'CbCr[A] 10bv 4:4:4[:4]"
imtc:v410:Shr5 "Sheer Y'CbCr[A] 10bv 4:4:4[:4]"
imtc:yuv2:Shr0 "Sheer Y'CbCr 8bw 4:2:2"
imtc:yuv2:Shr4 "Sheer Y'CbCr 8bw 4:2:2"
imtc:yuvs:Shr0 "Sheer Y'CbCr[A] 8bv 4:2:2[:4]"
imtc:yuvs:Shr3 "Sheer Y'CbCr[A] 8bv 4:2:2[:4]"

SYNOPSIS OF CURRENT STANDING: Pardon my crying Chicken Little with this one earlier, but I now think it breaks down like this:

1.) After Effects CAN render 10 bit QuickTime files with a proper 10 bits/channel of information, BUT....

2.) You have to be SURE you have your After Effects properly configured. This may well mean running BMD installers AFTER installing After Effects, and/or having to tweak the Adobe After Effects Prefs file by hand. Not ideal, but functional. Therefore...

3.) If you are going to use After Effects for 10 bit QuickTime renders, it behooves you to test the codec of choice on THAT PARTICULAR MACHINE and make sure it is working right. That means that it will BOTH let you select Trillions, AND that it actually renders out 10 bits worth of detail. To test, simplest way:
-make a 1024x10 pixel comp, 1/2 second or so long, 23.98 or 29.97 fps (really only has to be one frame, but I feel better rendering a few frames to make sure all is well)
-make a new solid, comp size
-Apply the Ramp filter (make sure your version is at least 16bpc capable), pure black to pure white, from 0,0 to 1024,0
-make sure project is set to at least 16 bpc
-render out to the codec of choice, being sure to select Trillions in the Output Module.
-and remember, just because it LETS you select Trillions, doesn't mean the file you make will BE Trillions (10 bits of detail)
-reimport that file you made, adjust the Info Palette units to a 10 bit scale (0-1024), and slowly move your cursor over the rendered file (drop in a comp if you have to) at at least 100% zoom, and make sure for each horizontal pixel you pass the cursor over moving from left to right the color value increments one value at a time, not in jumps of 4.

Simply hiding (via Stuffit compression) codecs, loading them in and out, doesn't make After Effects happy. After swapping out AJA for BMD codecs and relaunching After Effects to reopen my test file (that uses BMD codecs, and has BMD codecs queued for output) after hiding the BMD codecs and loading the AJA codecs LOCKED UP After Effects. So be careful

Codecs I know work in 10 bits:
-BlackMagic 10 bit RGB
-BlackMagic 10 bit 4:2:2

AJA codecs - I think I still don't have something quite set up right, I can do RGB+Alpha in Trillions for the AJA 10 bit log RGB codec, but when I switch it to plain RGB (no alpha), I only have Millions of colors avaiable. Actually, that applied to all the AJA codecs. Checking rendered output, somethin' weird happened, I'll have to play more to say definitively.

Apple 10 bit 4:2:2 uncompressed - did some funny gamma/ramping stuff to the image, I'm going to install Final Cut 5.1.1 tonight and retest.

...and then it's time for Sunday Night HBO, so I'm out. If you're hanging on the edge of your seat to see how all this turns out.....suffer, bitches! (gotta say that, The Wire is on....)

: )

-mike out