Audio Sync Issue after Deinterlacing and Converting to MP4

[cholla]

I'm believe Sharc gave me the MCDegrainSharp.avsi or a link to it.
Mine is for 32-bit avisynth plus.

It goes in the "plugins+" folder.

[Bermuda1]

Deleted - see post #30

[Bermuda1]

@Sharc
So I made a few changes to your script. I could not find the SmoothLevels on the avisynth website so I used SmoothAdjust instead - seems to do the exact same thing. I also set the QTGMC preset to slower. Otherwise I kept everything as per your script and the result is fantastic I must say. Thanks a lot! See video attached.

AviSource("Beethoven_Klaviersonate-short.avi").trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="Slower", EdiThreads=2)
crop(12,0,-12,-8)
SmoothAdjust_SmoothLevels(16,1.0,235,0,235)
bilinearresize(width/2,height).bicubicresize(width,height)
MCDegrainSharp()
converttoYV12()
#now encode x264 using -SAR 12/11
Prefetch(2)

[rgr]

Use AviSource() for .avi files, not FFMPEGSource2

Eventualy use the original audio (which is untouched anyhow) in the final output, with AudioDub() command

ffms2 works great with DV-AVI. Better than LibavVideoSource, but I don't remember what was wrong with the latter.

[lollo]

Not as good as installing a DV codec and using AviSource() in my experience

[rgr]

Not as good as installing a DV codec and using AviSource() in my experience

As you can see from this thread -- not necessarily.
And I've already converted a lot of DV-AVI tapes using ffms2.

[Sharc]

@Sharc
So I made a few changes to your script. I could not find the SmoothLevels on the avisynth website so I used SmoothAdjust instead - seems to do the exact same thing. I also set the QTGMC preset to slower. Otherwise I kept everything as per your script and the result is fantastic I must say. Thanks a lot! See video attached.

AviSource("Beethoven_Klaviersonate-short.avi").trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="Slower", EdiThreads=2)
crop(12,0,-12,-8)
SmoothAdjust_SmoothLevels(16,1.0,235,0,235)
bilinearresize(width/2,height).bicubicresize(width,height)
MCDegrainSharp()
converttoYV12()
#now encode x264 using -SAR 12/11
Prefetch(2)

Just to mention a difference: For no obvious reason your variant (left) denoises more but blurs more than mine (right) of post#29. See for example the plants in the background. The compromise is a matter of personal preference though, not to argue.

Edit: I see you were using a different (blurrier?) source file. That may be the reason for the difference in sharpness/details.

[lollo]

As you can see from this thread -- not necessarily.

Where did you see in this thread that ffms2 is better? Did I miss something?

And I've already converted a lot of DV-AVI tapes using ffms2

Sure. I just dislike its many versions around, leading different results.
Interesting reading here: https://forum.videohelp.com/threads/382528-How-to-achieve-best-results-decompressing-d...fore-restoring

[Bermuda1]

@Sharc
So I made a few changes to your script. I could not find the SmoothLevels on the avisynth website so I used SmoothAdjust instead - seems to do the exact same thing. I also set the QTGMC preset to slower. Otherwise I kept everything as per your script and the result is fantastic I must say. Thanks a lot! See video attached.

AviSource("Beethoven_Klaviersonate-short.avi").trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="Slower", EdiThreads=2)
crop(12,0,-12,-8)
SmoothAdjust_SmoothLevels(16,1.0,235,0,235)
bilinearresize(width/2,height).bicubicresize(width,height)
MCDegrainSharp()
converttoYV12()
#now encode x264 using -SAR 12/11
Prefetch(2)

Just to mention a difference: For no obvious reason your variant (left) denoises more but blurs more than mine (right) of post#29. See for example the plants in the background. The compromise is a matter of personal preference though, not to argue.

Edit: I see you were using a different (blurrier?) source file. That may be the reason for the difference in sharpness/details.

Hey Sharc,
It’s the same source file, I just renamed it. Since there are only two differences in our code, I can only imagine it’s either because I used QTGMC “slower” or because SmoothAdjust works different than SmoothLevels. I will try using the “fast” preset that you used. See how it comes out. Will update later.

[Bermuda1]

So I used QTGMC preset=Fast now. That did not change anything. Still very blurry compared to your version. @Sharc: can you provide the .dll or .avsi file for "SmoothLevels" that you used please? I am curious if its the "SmoothAdjust" that I am using, which is causing the blurry effect.

[Sharc]

Still very blurry compared to your version. @Sharc: can you provide the .dll or .avsi file for "SmoothLevels" that you used please? I am curious if its the "SmoothAdjust" that I am using, which is causing the blurry effect.

v3.20 is available from here:
http://avisynth.nl/index.php/SmoothAdjust

or here the plugins.
In your script simply use it as SmoothLevels(......). It's included in the .dll

Edit:
Maybe it's my bad and I picked the file which used less horizontal downscaling for the dotcrawl suppression. You can experiment with it.
I think this would explain the mystery. Sorry for the confusion. Try with:

.....
bilinearresize(480,height).bicubicresize(width,height)
.....

[Bermuda1]

Still very blurry compared to your version. @Sharc: can you provide the .dll or .avsi file for "SmoothLevels" that you used please? I am curious if its the "SmoothAdjust" that I am using, which is causing the blurry effect.

v3.20 is available from here:
http://avisynth.nl/index.php/SmoothAdjust

or here the plugins.
In your script simply use it as SmoothLevels(......). It's included in the .dll

Edit:
Maybe it's my bad and I picked the file which used less horizontal downscaling for the dotcrawl suppression. You can experiment with it.
I think this would explain the mystery. Sorry for the confusion. Try with:

.....
bilinearresize(480,height).bicubicresize(width,height)
.....

I now use SmoothLevels and changed the bilinearresize operation. Still mine is more blurry. What Ratefactor did you use? I had it set to 17.

Edit: reuploaded both the avi and the mp4 using this script:

AviSource("Beethoven_Klaviersonate-short.avi").trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="Fast", EdiThreads=2)
crop(12,0,-12,-8)
SmoothLevels(16,1.0,235,0,235)
bilinearresize(480,height).bicubicresize(width,height)
MCDegrainSharp()
converttoYV12()
#now encode x264 using -SAR 12/11
Prefetch(2)

[Sharc]

Your source is TFF not BFF. Change it in the script.

[Bermuda1]

Your source is TFF not BFF. Change it in the script.

Yes, you are correct, of course. I did create the above mp4 using TFF though. Stupid me just copied the incorrect script into here first. Edited in my above post to reflect the (correct) script that I used.

[Sharc]

The versions are identical now. The subtle differences are due to the x264 settings. My encoding commandline:

ffmpeg -i script.avs -c:a aac -c:v libx264 -preset slow -level 41 -bf 3 -g 50 -crf 18 -sar 12/11 "out_480.mp4"

As I wrote you may want to play with the downscale width (360 ....720) to find your compromise for the dotcrawl suppression.

Or you could try a different denoiser, for example:

AviSource(......).trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="fast")
crop(12,0,-12,-8)
Smoothlevels(16,1.0,235,0,235)
TemporalDegrain2(degrainTR=3)
converttoYV12()

[Bermuda1]

The versions are identical now. The subtle differences are due to the x264 settings. My encoding commandline:

ffmpeg -i script.avs -c:a aac -c:v libx264 -preset slow -level 41 -bf 3 -g 50 -crf 18 -sar 12/11 "out_480.mp4"

As I wrote you may want to play with the downscale width (360 ....720) to find your compromise for the dotcrawl suppression.

Or you could try a different denoiser, for example:

AviSource(......).trim(5,0).converttoYV16(interlaced=true)
assumeTFF()
assumefps(25)
QTGMC(preset="fast")
crop(12,0,-12,-8)
Smoothlevels(16,1.0,235,0,235)
TemporalDegrain2(degrainTR=3)
converttoYV12()

I tried different versions now. I prefer the 480 downscale width. Using 360 or width/2 is just a bit too blurry / too little detail left. Thanks for the recommendation!

On a side note: I also used FFMPEG instead of Virtual Dub for encoding. However, in order to get SAR 12/11 I had to amend your command line to include "-vf "setsar=sar=12/11"". Not sure why, but simply using -sar 12/11 did not work for me.

Edit:
I used TemporalDegrain2. Nice results, but the encoding now takes ages compared to using MCDegrainSharp.

[rgr]

As you can see from this thread -- not necessarily.

Where did you see in this thread that ffms2 is better? Did I miss something?

https://forum.videohelp.com/threads/416050-Audio-Sync-Issue-after-Deinterlacing-and-Co...P4#post2752153

Not as good as installing a DV codec and using AviSource() in my experience

I don't know if anything has changed, but Vegas codecs have always been considered the best for DV.
And I trust ffmpeg codecs more than random ones.
Also "ffmpeg/ffms2 defaults to error concealment with DV bitstream errors (there are always some)"

* - I remembered why I don't use LWLibavSource - it didn't see the audio.

[rgr]

However, in order to get SAR 12/11 I had to amend your command line to include "-vf "setsar=sar=12/11"". Not sure why, but simply using -sar 12/11 did not work for me.

You can add PAR settings to your Avisynth script:

propSet("_SARNum", 12)
propSet("_SARDen", 11)

And to the ffmpeg line:
ffmpeg.exe -i in.file -avisynth_flags +all out.file

[Sharc]

As you can see from this thread -- not necessarily.

Where did you see in this thread that ffms2 is better? Did I miss something?

https://forum.videohelp.com/threads/416050-Audio-Sync-Issue-after-Deinterlacing-and-Co...P4#post2752153

There has been a timestamp glitch in that clip at the beginning, see post#22. So it is debatable whether AviSource() of ffms2() behaved more correctly or "better", e.g. with respect to keeping A/V in sync.
A problem with ffms2 is indeed the various versions floating around, behaving differently, for example with RFF flags.

[lollo]

As you can see from this thread -- not necessarily.

Where did you see in this thread that ffms2 is better? Did I miss something?

https://forum.videohelp.com/threads/416050-Audio-Sync-Issue-after-Deinterlacing-and-Co...P4#post2752153

There has been a timestamp glitch in that clip at the beginning, see post#22. So it is debatable whether AviSource() of ffms2() behaved more correctly or "better", e.g. with respect to keeping A/V in sync.
A problem with ffms2 is indeed the various versions floating around, behaving differently, for example with RFF flags.

That's what I have concluded as well. (BTW, AFAIK the Vegas codec does not exist, it's a the Sony codec)

[rgr]

(BTW, AFAIK the Vegas codec does not exist, it's a the Sony codec)

The Sony codec is only available in Vegas -- there is no separate one (except for some exes for XP).
When using AviSource, the default(?) codec will be the Microsoft one, which has a "never-use-it, worst-possible" reputation.

[lollo]

The Sony codec is only available in Vegas -- there is no separate one (except for some exes for XP).

I remember at that time there were many (hacked) version of the Sony DV codec available

When using AviSource, the default(?) codec will be the Microsoft one, which has a "never-use-it, worst-possible" reputation.

According to the "codec merit", or removing some, you can specify whatever you want (Canopus, Sony, Panasonic, Cedocida, MainStream, etc)