No, Color Finesse always works in 32-bit floating point internally.
If you have an 8- or 16-bit project, Color Finesse will convert the pixels to 32-bit, do the color correction, then convert back to 8- or 16-bit. This way you only get any rounding error once—during the conversion back to 8- or 16-bit—rather than on each intermediate color correction operation. The difference is quite visible.
That said, if you have a 32-bit project you eliminate the overhead of those conversions, plus get the benefit of pixel values above 1.0.
Could be an issue with the initial conversion to DNxHD. The artifacts look like block edges of the chroma undersampling in compressed footage. I don't see how this could happen in camera, as H.264 is 8bit to begin with, but if there is misalignment between luma and chroma and a higher bit-depth CoDec responds to it by encoding it as non-quantized values, it would then show up.... Probably just some switch/ option you need to tick to avoid it. Also note that all Avid CoDecs do not respond well to having unnecessary Alpha channels attached when that Alpha contains nothing but solid fill....
I also saw this behaviour mysteriously appear recently.
I don't remember doing any After Effects or Avid updates at all, and I say mysteriously because I literally have footage I rendered in DNxHD a few months ago that does not display the problem. Really quite odd.
Anyway, here are more detailed screenshots:
Notice, if you change AE to 8bpc, problem goes away:http://yfrog.com/z/h4tvv4nj
Super weird, right? It's not something that's actually on the footage--the footage looks great and renders fine in Avid and Premiere.
On Mylenium's queue, I tested out exporting a bunch of different switches before bringing into AE. Here's the results:
One note, this is with AE 16bpc mode on, I forgot to not crop that out.
The run down? The third image down is the settings I've always used. Has the black artifacting in the highlights.
For the image above that (2nd), I tried just changing Adobe Media Encoder to "64-bit" instead of leaving it on "24-bit", making sure "use maximum bit rate" is checked. It still has the artifacting, though a different pattern.
On the bottom image, I went into DNxHD's settings and changed from "709" to "RGB". From everything I've read, this is a no-no and you're supposed to use 709 across the boards. But it solved the "black things" issue. And, as you can tell, didn't seem to affect any of the darks/lights on my footage. I'm not an expert and I don't know what this switch "really does" and what broadcast spec. stuff this might cause problems to do.
On the top image, I tried not rendering out with DNxHD 175 X, just DNxHD 175, which is 8-bit and not 10-bit. On the shadow area I annotated, this as expected, introduces some nasty banding problems, although it also solves the black spots even when AE is in 16bpc.
The one other thing I did notice is that, when using the "RGB" check instead of "709", there is a very slight banding introuduced in the same area. It's hard to see, but it's there, where it isn't on the 709 exported material.
These are my findings. Maybe some other expert here has a better idea of why and ideas to solve the problem. My questions:
- what exactly choosing 709 or RGB does to the image--I can't tell. I've been testing out Cineform, partially due to this black spots issue, and it doesn't have 709 or RGB options, along with every other codecs I can think of. In fact, it's only there in Avid and no one in Avid seems to actually use RGB.
- why there is 24-bit 32-bit 48-bit 64-bit options when encoding in AME, and what they do when choosing them (I can't tell any differences). And, if I leave "maximum bit rate quality" off, does that mean it's not using any of these and rendering 8-bit like the tooltip says? Or is it rendering 24-bit which is the lowest you can pick?
- why even DNxHD encoded without the X (10-bit), still shows up at "Trillions of Colors" in After Effect's info? More specficially, how does the millions colors/trillions colors relate to bit depth?
Hope this helps anyone running into this problem.
I'm having the same problem.
I am rendering from DVCPRO through an AE - 16 bit - rec 709 timeline to 10 bit QT-DNxHD (No problems at this point). I then edit in PP. No signs of problems at this point either. From there I export via AME to AVI, H264 and Mp2 DVD. This is where it gets bad. All clipped highlights, coming from PP through AME, have the black spots in them.
I only get this when DNxHD has been used in the process. No other codecs give this problem to date.
Has anyone found an answer?
Did you click further up into the QuickTime codec settings and select RGB instead of 709? If I'm on 709 with 10-bit, I get highlight issues on the next encode with certain codecs. If I just change that initial encode to RGB (and keep it RGB though any & all intermediates), highlight issues be gone.
You can check it quickly by bringing it back into a 16bpc or 32bpc After Effects comp and seeing if the highlight issue is there on the footage. In 8bpc no issues.
You could try some other codecs just to see and definitely keep off "use maximum render quality" if your source footage didn't have RGB checked--as that should force it to encode into 8-bit regardless... right? I'm only assuming here.
Additionally I'm getting a similar but different artifacting whenever I use "use maximum render quality" or whatever it is called. I don't think this one is DNxHD specific.
All in all pretty annoying to put up with and hoping at some point Avid & Adobe address the issues, but not holding my breath, it doesn't seem like many people DNx 10-bit workflow in Premiere specifically (Media Composer's output is fine) so there's not a lot of feedback on it to encourage fixes.
I tried changing to RGB in the output settings dialog, and the problem went away in both 16 and 32 bit. It's early days yet, but it looks as though this might have fixed the problem. Thanks for the advice! I thought I'd already tried this but obviously I hadn't. Go figure
I've had this problem also. It's been an issue for quite some time, and it's very annoying when you're trying to have a 10-bit workflow between Avid and AE.
There must be some error in the way AE interprets DNxHD. It doesn't matter what program originally created the DNxHD file. Even if you encode in AE you get this problem. I've used DNxHD as proxy codec a few times and have had this problem.
Exporting from Avid using the RGB setting solves the problem, but only by clipping off the high whites. I have a few times applied a colour correction inside Avid to reduce the levels to within 16-235 before exporting as RGB to avoid the clipping.
I'm pretty sure the issue lies in any non-0 9th or 10th bit whenever the other 8 bits are all 1s. Ie the pixel is fully bright. When you export as RGB the values get clipped and the problem goes away.
I think your assessments are correct BC. Truthfully between these highlight issues, and needing to deal with material beyond 1080p at times, we're using Cineform more and more here for our cross platform work. It's a bit pricey and licensing is rather strict, but in general it's been fairly rock solid playing intermediary.