Intermediate XQ codec: Not true 10 bit? (plus an audio request)

Greg-Kintz wrote on 10/4/2019, 2:10 AM

I've been doing tests with the various Vegas intermediate codecs, and specifically the XQ format which in the Apple ProRes world world offers a 12 bit depth, vs the 10 bit of most of the other variations. (Some proxy LT variations can be 8 bit) Of course the Magix intermediate codec is not true ProRes, but is supposed to somewhat mirror it for the goal of delivering a compatible product world wide that adheres to certain standards.

I've used a 16 bit test pattern in the 32 bit mode and all non-proxy versions appear to be 10 bit, except for the highest quality XQ render setting, which oddly looks worse. As in, the XQ render appears to be 8 bit. It should look the best.

Further confirmation may be found in Adobe After Effects. When loading in these various clips, the properties for each showed "trillions of colors" except for the XQ render, which shows as "millions of colors".

 

----

..And ever so slightly off topic, but dealing with the Magix intermediate codec: If developers are reading, could an option be provided to include multi-channel PCM audio, vs the current stereo PCM only implementation? Apple ProRes allows for this option, and there's a LOT of material that needs to be delivered as 5.1.

Comments

Musicvid wrote on 10/7/2019, 8:19 PM

Seems like a lot of guesswork. Comparing static file sizes may give more clues as to whether bit preservation vs. decimation is going on.

Greg-Kintz wrote on 10/9/2019, 9:38 PM

It's not guesswork that the 16 bit test pattern looks worse in XQ vs HQ or 444. It's not guesswork that AE shows millions of colors for XQ but shows trillions for 444 and HQ.

A file's static size can show less compression is being done, but it is not a determination of if a codec is preserving true 10 bit resolution. That would be guesswork. Unfortunately there have been previous codecs that have claimed 10 bit resolution but merely did 8 and essentially padded to 10. Hence my tests and my post.

Former user wrote on 10/10/2019, 6:52 AM

@Greg-Kintz The following may, or may not be relevant to your main query, its purely about how well a codec rendered out holds up compared to the source, quality wise.

I previously did some quality comparisons between different codecs, including Magix Intermediate 422HQ etc, but didn’t include the XQ one, still haven’t as its probably unnecessary.

So I just did a quick test using ffmpeg, for SSIM ALL and PSNR Avg. Higher is better ...

422 HQ ... SSIM ALL .999352 (31.881710) PSNR Avg 56.533286

XQ .......... SSIM ALL .999450 (32.598767) PSNR Avg. 58.820011

So quality wise the XQ file comes in better than the HQ one.

The following table shows how well the HQ file already does compared to others.

 

 

Musicvid wrote on 10/10/2019, 7:39 AM

@Greg-Kintz

I've done quite a lot of testing over the years, and yes, size matters. True 10 bit occupies about 30% more space than either its exact 8 bit counterpart, or 8 bit miswrapped as 10 bit, which is not that uncommon.

Rather than say, "looks worse," post your quantified test results and conditions, along with histograms, of course.

The biggest problem we face in evaluating such observations is the tendency to misattribute outcomes, when the test conditions may be to blame.