Quality issues with Mainconcept/Magix AVC encoder?

fifonik wrote on 2/11/2018, 3:50 PM

Hi,

I used to have rendering issues with Mainconcept AVC encoder long time ago in Vegas Pro 12 that I reported to Sony. The Mainconcept encoder in VP11 was not affected so I used to replace VP11 mainconcept encoder dll in VP12 for a while and then I switched to Sony AVC encoder. Never worried about Mainconcept encoder since.

However, when I bough VP15 I found that I cannot use Sony AVC encoder any longer as it does not play nice with 32-bit video levels pixel format. So I decided to to give Magix AVC encoder another try.

I opened my old project from VP12 era where the glitch was extremely visible in VP13 and VP15 and rendered it with Mainconcept/Magix AVC (CPU Only/Mainconcept AVC mode as I have AMD graphics card). The glitch is still there.

I simplified the project and now it has only one media file (m2ts, ~50MB, 1080-50p from Panasonic TM900) and no filters. The project in 8-bits pixel format and render quality set to Best.

I use default Mainconcept/Magix Internet HD 1080p profile and only change framerate from 29.xxx to 25 or 50.

Then in rendered video I have very pixelated frames. Not all of then, only few, but they are very visible as when the file is played some areas looks like they are 'pulsing'. If stopped and watched by frame some frames are extremely wrong.

If I use variable bitrate mode -- the bitrate of resulting video is very low and not matched to specified average bitrate. If I use constant bitrate mode -- the bitrate of resulting video matched the bitrate I specified, but these very pixelated frames are still there.

Mainconcept encoder included in VP13 and VP15 (after changing Internal setting) are also affected. I was not able to check if mainconcept encoder dll from VP11 still do the trick as I do not have VP11 installer any longer and do not know where I can download it.

When I render using Sony AVC or with x264 encoders (using frameserver) and there are no such issue. My hardware & OS completely changed since than.

So, here some details.

Source file media info:

General
ID                                       : 1 (0x1)
Complete name                            : J:\Video\VegasMainconceptBug\2012-11-22_095841.m2ts
Format                                   : BDAV
Format/Info                              : Blu-ray Video
File size                                : 41.6 MiB
Duration                                 : 12 s 950 ms
Overall bit rate mode                    : Variable
Overall bit rate                         : 27.0 Mb/s
Maximum Overall bit rate                 : 28.0 Mb/s

Video
ID                                       : 4113 (0x1011)
Menu ID                                  : 1 (0x1)
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4.2
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, RefFrames               : 4 frames
Format settings, GOP                     : M=3, N=24
Codec ID                                 : 27
Duration                                 : 12 s 920 ms
Bit rate mode                            : Variable
Bit rate                                 : 25.5 Mb/s
Maximum bit rate                         : 26.0 Mb/s
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate                               : 50.000 FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.246
Stream size                              : 39.3 MiB (94%)

Audio
...

Project settings, Magix AVC CBR, Magix AVC VBR, Sony AVC encoders properties (orange -- name of default profile used, red -- fields that I modified):

Frame1. Look at grass and these red things (Magix CBR, Magix VBR, Sony AVC, x264):

Frame2. Look at face and T-shirt (Magix CBR, Magix VBR, Sony AVC, x264):

Link to zip file with the project & source media and screenshots (~100MB):

https://www.dropbox.com/s/1qabfl40ti0beq1/VegasMainconceptBug.zip?dl=0

Link to zip file with the project, source media, screenshots and rendered medias (~250MB):

https://www.dropbox.com/s/94gr6lh19rna41z/VegasMainconceptBugFull.zip?dl=0

Comments

fifonik wrote on 2/12/2018, 5:25 AM

Additional information & observations:

- Render using VP13 and VP15 Mainconcept AVC (after enabling it in Internal settings) gives results similar to Magix AVC.

- Render using VP11 Mainconcept AVC does not have such issue

- Now I have AMD CPU & graphics card so no NVIDIA or QuickSync options available

- I tried to further modify Magix AVC encoder settings by selecting High profile -- no big changes.

- Rendering using Sony AVC in VP15 gives result similar to VP13 SonyAVC

- Sony AVC gives slightly blurred, but not so "pixelated" results with even smaller bitrate (for some reason it does not use all bitrate specified in encoder settings).

fifonik wrote on 2/12/2018, 4:16 PM

Thanks Cornico for looking at this.

Can you point me out and specify on the next screenshot from that timeline where to look for your seen "pixelation"?

On the arm, face, T-shirt and helmet. They are less visible as screenshots were resized. I'm opening two images in 100% zoom with my picture viewer and switching from one to another.

I uploaded Frame2 (SonyAVC vs Magix AVC VBR) to screenshot comparison:

http://www.screenshotcomparison.com/comparison/131707

Scroll to the boy, and hover mouse pointer over picture to see Magix frame. Move mouse out to see Sony AVC.

In dynamic (when you watch the video) issues on these red things is very visible as they are 'pulsing' (not sure how to explain it better, English is my second language, sorry). This is how I originally spotted them.

In my original project the situation was even worse. It is possible to make it worse in the test project: just move the source footage to 00:02:30 and fill in the gap with something "media generated" (Solid colour, checkerboard. In my real project it is other footages here where there is no grass that is always hard for encoding).

I'll definitely check the project you uploaded at home.

However, could you advice, what project settings you mentioned? I have not noticed any differences in you project settings picture other than Deinterlace method that is not relevant here as there is no interlacing involved at all. Did I miss anything?

I created the demo project in VP11. The reason -- I'd like to be able to open it in different VP versions (VP11, 12, 13, 15) for checking. However, when I tested, I created the similar projects in VP15 with the same results.

P.S. I saved images from browser preview and they are smaller than I uploaded. Looks like the forum modifying these images. So for comparison the screenshot from the archive I uploaded yesterday should be used.

Musicvid wrote on 2/12/2018, 7:19 PM

Was hardware assist on or off during your tests?

fifonik wrote on 2/12/2018, 11:04 PM

Vegas Settings:

GPU acceleration is switched ON. I though about switching it OFF, however decided not to do this as there are no quality impact when SonyAVC or frameserver used.

Encoders:

In VP13 encoders setting 'Render using CPU Only' was used (the option is available on encoder properties).

In VP15 there is no such option on encoder properties as I have AMD CPU & GPU. Also, there are no profiles for NVENC/QSV available.

fifonik wrote on 2/13/2018, 2:18 AM

2 Cornico:

I downloaded and checked your renders. The issue is still there. You can compare Sony AVC frame 491 vs MC AVC frame 490 (I do not know why the frames slightly shifted).

Then I looked at your QS and NVENC results and found that QS is even worse than MC AVC, but NVENC is the best from them (however, I'd say they are worse than Sony AVC on lower bitrate).

Archive with the screenshots from your renders:

https://www.dropbox.com/s/k4fm64e2vwigutm/FromCornico.Frame%20490.zip?dl=0

fifonik wrote on 2/15/2018, 4:20 AM

Having fun with video analyzing...

Simplified project (only one fragment in project with absolutely no filters, transitions etc), rendered into VP15 Magix AVC/SonyAVC/x264 with slightly different encoder settings (so renders files at the end have the similar bitrate) and using MSU Quality Measurement Tool for analyzis. Interesting. Too many new things (PSNR, SSIM, VQMT etc). Something looks really funny, for example this 'blocking' metric:

P.S. One more time: Mainconcept AVC encoder included in VP11 does not have such issue.

fifonik wrote on 3/24/2018, 10:49 PM

Finally, I finished the encoders comparison (VP15 Magix AVC, VP15 Sony AVC, VP11 Mainconcept AVC, x264, all CPU only).

Here you are (~270MB):

https://www.dropbox.com/s/3ij52xkc9p3ex9q/VegasMainconceptBug2.zip?dl=0

In the archive you can find m2ts source media from camcorder Panasonic TM900, veg-project created in VP11, a few mp4-files rendered with different encoders, avs scripts that can be used to compare results using MSU Video Quality Measurement Tool, measured metrics (csv + png) created with the program, descriptions (RU + EN).

Download, unpack.

Then you have options:
1. Just watch rendered files (in dynamic of frame by frame) and also check included metrics (png files in screenshots folder)
2. Install MSU Video Quality Measurement Tool Free, AviSynth, LSMASHSource plugin. Run the MSU Video Quality Measurement Tool, oped prepared vqmt files (change paths to files in these files if required) and run Start button.
You can render the project using your favourite encoder (QSV, NV, HEVC), create avs script for these files (copy/paste existing avs file and change file name) and compare them as well.

A few metrics from the archive:

VP15 Magix AVC vs VP11 Mainconcept AVC

Blocking metric:

SSIM metric:

VP15 Magix AVC vs VP15 Sony AVC

Blocking metric:

SSIM metric:

VP 15 Magix AVC vs x264 (rendered from VP15 using frameserver and MeGUI)

Blocking metric

SSIM metric

fifonik wrote on 3/30/2018, 11:31 PM

Reported [Ticket#2018033117001059]

mintyslippers wrote on 4/12/2019, 4:07 AM

I thought I would add to this excellent post as your little package for testing is great. Attached are screenshots from my own renders on VP16. I couldnt get the quality tool to work as I dont have pro but visual was more than enough to confirm nothing has improved in this version.

Specs: AMD Ryzen 2700x, Nvidia RTX 2060 using 419.67 CRD Drivers

Uncompressed (all is fine)

VP 16 using Magix AVC and the Mainconcept encoder (blocky as hell)

VP 16 using Magix AVC and the NVENC encoder (better but still sub par)

VP 16 and Sony AVC (the benchmark IMHO)

VP 16 and Magix HEVC and NVENC (the best option for speed and quality. But still not Sony AVC levels)

Magix devs. I know you read these forums. We love Vegas. Can we get some input here? For footage which uses a higher bitrate than Sony to deliver MUCH lower quality does warrant some investigation. As a test I used the sony AVC but at a mere 10mbps. Over half that used for the Magix encoders.

I get the exact same results from my Canon C100 footage, 8bit

NickHope wrote on 4/12/2019, 12:36 PM

...VP 16 using Magix AVC and the Mainconcept encoder (blocky as hell)...

@mintyslippers How does it look if you check "Use deblocking filter"?

mintyslippers wrote on 4/12/2019, 12:47 PM

...VP 16 using Magix AVC and the Mainconcept encoder (blocky as hell)...

@mintyslippers How does it look if you check "Use deblocking filter"?

Not a single bit of difference. That's why I think it's not setting the parameters correctly. 2 pass is slower than Sony and still not as good.