what’s really going on in ProRes? (or 10bit 4:2:2: uncompressed)?

[Jeremy Garchow]

[josh litle] “There are definite and obvious color and gamma shifts between the uncompressed 10-bit 4:2:2 and the ProRes-compressed 10-bit 4:2:2. “

Wild guess here, but instead of using software or hardware that can handle this process, you simply exported out of Quicktime or used Quicktime Conversion?

[josh litle] “Basically creating bars in both uncompressed and ProRes formats, then exporting them to a standard TIFF format (uncompressed) and comparing them in Photoshop.”

You are getting way ahead of yourself, again, what is your process for this?

Jeremy

[Sascha Engel]

I experienced a similar phenomena: I exported a finished video clip – one in Uncompressed 10bit 4:2:2, in Uncompressed BlackMagic and in Appple ProRes HQ and all three of them looked very different.
I ended up choosing the Blackmagic codec for looking the most crispy, strongest blacks and best contrast.
But why? I still have NO clue!
Hope somebody can enlighten us!
🙂

Sascha

[Jeremy Garchow]

[Sascha Engel] “I exported a finished video clip – one in Uncompressed 10bit 4:2:2, in Uncompressed BlackMagic and in Appple ProRes HQ and all three of them looked very different. “

And how did you export? Quicktime Conversion?

[Alan Okey]

[josh litle] “I recently had a film color corrected in Scratch and rendered out to Quicktime Apple Uncompressed 10-bit 4:2:2 codec. I then made a ProRes HQ export of this movie. Guess what? They don’t look the same. There are definite and obvious color and gamma shifts between the uncompressed 10-bit 4:2:2 and the ProRes-compressed 10-bit 4:2:2. “

The most important piece of information that’s missing here is how you are monitoring these two files. Are you monitoring the output on a broadcast monitor by doing an A/B between the two different files in FCP, or are you looking at them in Quicktime player on your computer monitor?

[Sascha Engel]

I exported it from FCP as follow:

Edit -> Export -> Quicktime Movie (not Conversion!) -> then I adjusted the settings there to the Codecs I mentioned above.

[Jeremy Garchow]

[Sascha Engel] “then I adjusted the settings there to the Codecs I mentioned above. “

Yeah, that’s probably your problem.

I’d use Compressor to transcode with frame controls on (to ensure 10bit transcode).

Or

Make a new sequence with the appropriate codec and render, then export a movie @ current settings.

You have to do this the ‘right’ way, or FCP/QT won’t handle it properly.

Jeremy

[Josh Litle]

Well, yes I did export from Quicktime. So I will attempt some tests following the recommendations above.

Regarding the monitoring question, yes I am monitoring on the desktop. That is why I tried creating RGB Tiff files – so I could compare the processed images in the same format in the same app (Photoshop).

Note to Apple: The FCP / QT / Compressor apps really should all handle this stuff (e.g. codec, bit-depth, colorspace, transcode) identically. As users we should not have to know all of these obscure work-around for simple export to be handled correctly. Grump. I always assumed the QT engine was at the heart of all three apps.

josh
filmmaker

[Alan Okey]

[josh litle] “Regarding the monitoring question, yes I am monitoring on the desktop. That is why I tried creating RGB Tiff files – so I could compare the processed images in the same format in the same app (Photoshop). “

Your video files are not RGB. Uncompressed 10-bit and ProRes 422 are both YCrCb codecs, not RGB. By viewing them on a computer monitor in RGB colorspace, you’re at the mercy of Quicktime Player’s YCrCb to RGB conversion, plus Quicktime’s notorious gamma issues. The inconsistencies you see are most likely the result of how the files are being (incorrectly) displayed in RGB colorspace by Quicktime Player, not necessarily a problem with the video codecs themselves.

Keep in mind that the only way to accurately view and evaluate the color of your video files is on a proper broadcast monitor and/or hardware scopes fed by a proper video I/O device like an AJA Kona card. Trying to compare them in Quicktime Player on a computer monitor will not yield accurate results for a multitude of reasons.

[Josh Litle]

Thanks Alan,

I understand your point about the YCrBr to RGB conversion. I do think that since both files are being converted to RGB using the same app (quicktime player) that presumably they are undergoing the same conversion process (accurate or not); then when viewed as RGB TIFF file in the same App (Photoshop) they theoretically will look the same (if the original Quicktimes were the same). But that’s really academic, since it relies on visual comparisons and many different variables in the chain.

I performed a much more accurate comparison that didn’t involve display issues or the human eye: I imported both QT’s onto seperate layers and composited them using difference mode (in AE), which only displays the parts of the files that are different from one another. If the files are identical this will result in an absolute black image (no difference between the files). However in this case you could clearly see a strong “ghost” image (much more than normal compression artifacts). I also compared the two files inside of FCP using the built in Vectorscope and waveform tools. Again the values were clearly different. Both of these methods should (in theory) bypass any display or color managing by Quicktime, FCP or the operating system. They are (in theory) mathematical operations based on the actual pixel values present in the files.

I still need to try the various export methodologies outlined above. I will report back on that.

josh
filmmaker

[Sascha Engel]

Thanx. New thing learned. Will try it next time!