AVStoDVD Support Thread

[_MrC_]

AQ is still 1 by default because it is my personal taste. Feel free to change it if you want.

Bias is calculated by AVStoDVD at runtime using this formula: Bias = Int(VideoBitRate / 1000) x 10

Mike, if you are getting an error message when you do the FrameServing Check then there is a high probability that your directshow codec chain is messed up. Follow dave suggestion to use GraphEdit and report back.

Just another question: do you get errors when using other files than the "problematic" flv?




Bye

[Mike99]

Shortly after my last post I realized that I forgot to re-register FLVSplitter. After doing that HCenc seems to work OK. Hurray!!

However when doing a 1-pass, the HCenc box opens up & the Info line states “very low quantizer, bad Q estimation, please reconsider settings”. What does this mean?

When doing a 2-pass, the Info line states “bitrate set to 6000”.

Both times I used 6000 bitrate, 9000 max bitrate. And the FLV file was about 6 minutes.
For the 1-pass AQ=1 and for the 2-pass AQ=2. .

Both times when the HCenc box closed I got an AVStoDVD message:
“Warning! Output Video duration (00:00:00) is shorter than expected duration (00:05:52). Do you want to continue anyway?”

I selected yes and everything seemed to proceed OK, at least I got a working VIDEO_TS folder. The FLV file was about 6 minutes, but why does the warning state “Output Video duration (00:00:00)”?

Thank you MrC and davexnet for all your help.

[davexnet]

Mike, I can only say (at this point ) that I got the same message about the duration. I selected yes, and it completed
properly. For some reason, the tool that AVStoDVD uses to check the m2v when the encode finishes has an occasional problem.
(I think this is something MrC may have to comment on).

I have seen good m2v files receive this message, even when manually opening the file with mediainfo and the
information looking good.

For one-pass HCenc, do not worry about the "very low quantizer " message. This is just HCenc doing a pre-encode
test of the file trying to find out how to set itself to honor your selected bitrate in one pass as closely as possible.

[Mike99]

I was comparing some frame captures of QuEnc vs HCenc 1-pass vs HCenc 2-pass. So far it’s really a toss up when pixel peeping the 3 different versions.

However when I toggled between QuEnc & HCenc I noticed the picture jumped a bit. IOW they were not the same. Turns out the HCenc stills and video have a thin black line at the top & bottom. This also appears on the sample that davexnet did. Therefore the image is slightly shorter. Same content, just that it has to be shrunk a bit in order to fit within the black lines/borders.

The original FLV file does not have the lines. Maybe it does but it displays differently. I don’t know. It is 640x360 which is 1.7777:1 aspect ratio which is correct.

Any idea why black lines are being added?

[davexnet]

If you create a simple script and open it in virtualdub, you'll see dimension 640*368, 4 pixels of black
top and bottom. There are those mystery pixels. (Here's the script)
DirectShowSource("F:\Documents and Settings\Dave Xnet\My Documents\Downloads\Hulu - Peacemakers.flv", fps=23.976, audio=false)

However, I disabled the H264/AVC decoding in FFDshow and enabled/re-registered CoreAVC decodng instead
(which I already had installed in an inactive state) Now when you open the script in Virtualdub, the dimensions
are 640*360 and the black pixels, top and bottom, are absent.

I've never seen this before, FFDshow seems to be implicated. Perhaps this is a question for one of the other
sub-forums here at Videohelp.com. Someone can perhaps explain it.

[_MrC_]

For some reason, the tool that AVStoDVD uses to check the m2v when the encode finishes has an occasional problem.
(I think this is something MrC may have to comment on).

Yeah, I have. There's a small bug in the media reading routine. Will fix in 2.3.3 release (quite ready).



Bye

[Mike99]

davexnet,

If I understand this correctly, virtualdub indicates the FLV file is 640x368 with extra black pixels and CoreAVC indicates it is 640x360 with no black pixels. FWIW both MediaInfo and VLC indicate 640x360 and VLC plays with no black bars.

When I looked at the frame capture JPG images it appears the QuEnc version without the black lines is actually stretched a bit vertically. This is based on looking at the round Universal logo which appears at the beginning of episode #1 which I’m presuming really is round. I put a square box around it and the globe does not fit exactly. However the whole rectangular image is 1.78:1 AR.

Doing the same with an HCenc version the globe pretty much fits exactly. If I crop just the active part of the image (no black lines) the image is about 1.82:1 AR. Obviously not a huge difference but which is correct.

Would this have something to do with the 16x16 macroblocks of pixels? 360 is not a multiple of 16 but 368 is.

Do FLV files have to be multiples of 16? I don’t know how to do an FLV frame capture so I just measured the logo on the screen & it is slightly taller than wide, but that’s not a real accurate way to measure.

Maybe QuEnc is retaining the FLV height even though it appears to be slightly tall. Or if my measurements are off could QuEnc be stretching the 360 to 368?

Could HCenc be adding extra black pixels to make the 368?

[davexnet]

It may well have something to do with the mod16 sizes.
For now, if you just want to get a round it, disable the H264 decoding in the FFDShow video decoder config
and install CoreAVC ($10) or try the free Divx codec.
http://www.divx.com/
http://corecodec.com/products/coreavc

Another bonus of one of the above: faster decoding compared to FFDshow.

I'll open a post in one of the other forums to see if any of the other regulars have experienced this before.

[_MrC_]

there's something weird in that flv file: MediaInfo detects 23.976 fps, GSpot 24.013 and AviSynth 25 fps. I suspect that AviSynth is right.

Another trial to fix the out of sync issue.

0. Have flvsplitter and ffdshow installed
1. Go to 'Preferences'/'AviSynth' and select 'DirectShowSource' as AVS Source Filter
2. Add the flv
3. Go to 'Edit Title'/'AviSynth' and toggle off 'Auto AviSynth Script'
4. Assuming the input flv is "C:\file.flv", locate line

Video = DirectShowSource("C:\file.flv", fps=23.976, audio=false)

and change it into:

Video = DirectShowSource("C:\file.flv", fps=25, audio=false)

or:

Video = DirectShowSource("C:\file.flv", fps=24.013, audio=false)

5. Use the 'Preview Title Output' feature to check sync

Hope it helps



Bye

[Mike99]

It may well have something to do with the mod16 sizes.
For now, if you just want to get a round it, disable the H264 decoding in the FFDShow video decoder config
and install CoreAVC ($10) or try the free Divx codec.
http://www.divx.com/
http://corecodec.com/products/coreavc

Another bonus of one of the above: faster decoding compared to FFDshow.

I'll open a post in one of the other forums to see if any of the other regulars have experienced this before.

When I used QuEnc the ffdshow icon never appeared in the lower task bar so I'd have to presume that it was not running. At least that how it was without FLVSplitter. So I could just use QuEnc which would also be a work around. I'll have to see how it works with FLVSplitter on the system.

It would be nice to know if adding the black lines corrects the AR or makes it wrong.

Again looking at the short conversions I've done it's a toss up which looks better, QuEnc, HCenc 1-pass, or HCenc 2-pass. But I'll play around with some different settings.

[Mike99]

there's something weird in that flv file: MediaInfo detects 23.976 fps, GSpot 24.013 and AviSynth 25 fps. I suspect that AviSynth is right.

Another trial to fix the out of sync issue.

0. Have flvsplitter and ffdshow installed
1. Go to 'Preferences'/'AviSynth' and select 'DirectShowSource' as AVS Source Filter
2. Add the flv
3. Go to 'Edit Title'/'AviSynth' and toggle off 'Auto AviSynth Script'
4. Assuming the input flv is "C:\file.flv", locate line

Video = DirectShowSource("C:\file.flv", fps=23.976, audio=false)

and change it into:

Video = DirectShowSource("C:\file.flv", fps=25, audio=false)

or:

Video = DirectShowSource("C:\file.flv", fps=24.013, audio=false)

5. Use the 'Preview Title Output' feature to check sync

Hope it helps



Bye

With all the different fps readings how do we know for sure what the FLV file really is? FWIW previously I checked about 10 TV programs & movies on Hulu and they were either 23.976 fps or 29.97 fps. I'm certainly not challenging anyone & anything is possible, but I do question why a North American TV show would be made in 25 fps.

[_MrC_]

With all the different fps readings how do we know for sure what the FLV file really is? FWIW previously I checked about 10 TV programs & movies on Hulu and they were either 23.976 fps or 29.97 fps. I'm certainly not challenging anyone & anything is possible, but I do question why a North American TV show would be made in 25 fps.

I cannot answer to you. Maybe you should ask to North American broadcasting companies. Or to the guy that encoded that flv.

Recently I have down'ed a flv trailer detected by Mediainfo at 23 fps, by GSpot at 24.013 fps, by AviSynth at 25 fps. After "standard" DirectShow decoding it was out of synch. Using the explained workaround I found that correct framerate was 24.013 fps (thumb up to GSpot) because, using that setting in DirectShowSource (fps=24.013), output was in synch.

I have tried to give you a possible solution. It is up to you to try it.



Bye

[Mike99]

With all the different fps readings how do we know for sure what the FLV file really is? FWIW previously I checked about 10 TV programs & movies on Hulu and they were either 23.976 fps or 29.97 fps. I'm certainly not challenging anyone & anything is possible, but I do question why a North American TV show would be made in 25 fps.

I cannot answer to you. Maybe you should ask to North American broadcasting companies. Or to the guy that encoded that flv.

Recently I have down'ed a flv trailer detected by Mediainfo at 23 fps, by GSpot at 24.013 fps, by AviSynth at 25 fps. After "standard" DirectShow decoding it was out of synch. Using the explained workaround I found that correct framerate was 24.013 fps (thumb up to GSpot) because, using that setting in DirectShowSource (fps=24.013), output was in synch.

I have tried to give you a possible solution. It is up to you to try it.



Bye

I apologize if I sounded argumentative, that was not my intention. I’m willing to try things but I also have questions because I want to learn what is going on. A lot of this is new to me.

I do not have sync issues when using QuEnc with A2DSource & without FLVSplitter, or HCenc with DirectShowSource and FLVSplitter. Both create an output file with 23.976 fps. This may be wrong but they look better than 25 fps.

I tried QuEnc with A2Dsource & FLVSplitter. This created a 29.97 fps file and the audio was out of sync.

I then tried QuEnc with DirectShowSource and FLVSplitter. This created a 23.976 fps file and the audio was in sync. However this also creates the thin black lines/borders and the top & bottom.

I will try the 24.103 fps next.

[Mike99]

If you create a simple script and open it in virtualdub, you'll see dimension 640*368, 4 pixels of black
top and bottom. There are those mystery pixels. (Here's the script)
DirectShowSource("F:\Documents and Settings\Dave Xnet\My Documents\Downloads\Hulu - Peacemakers.flv", fps=23.976, audio=false)

However, I disabled the H264/AVC decoding in FFDshow and enabled/re-registered CoreAVC decodng instead
(which I already had installed in an inactive state) Now when you open the script in Virtualdub, the dimensions
are 640*360 and the black pixels, top and bottom, are absent.

I've never seen this before, FFDshow seems to be implicated. Perhaps this is a question for one of the other
sub-forums here at Videohelp.com. Someone can perhaps explain it.

I had someone make a 640x360 FLV file with a plain background and a large square. I figured I’d look at the image on my HDTV and measure the square to see it was really square.

Then I would convert to MPEG/DVD format using HCenc and see what happens to the square’s vertical dimension after HCenc adds the black lines. IOW which version is really correct. Well HCenc did not add black lines.

Rhetorical question I guess, why does HCenc add black to the Hulu file but not the homemade one? The homemade file is Video format VP6 per MediaInfo, and I’m presuming this may affect the conversion.

[davexnet]

Do the test I did above with the simple script and Virtualdub and see for yourself if the black lines are there.

[Mike99]

With all the different fps readings how do we know for sure what the FLV file really is? FWIW previously I checked about 10 TV programs & movies on Hulu and they were either 23.976 fps or 29.97 fps. I'm certainly not challenging anyone & anything is possible, but I do question why a North American TV show would be made in 25 fps.

I cannot answer to you. Maybe you should ask to North American broadcasting companies. Or to the guy that encoded that flv.

Recently I have down'ed a flv trailer detected by Mediainfo at 23 fps, by GSpot at 24.013 fps, by AviSynth at 25 fps. After "standard" DirectShow decoding it was out of synch. Using the explained workaround I found that correct framerate was 24.013 fps (thumb up to GSpot) because, using that setting in DirectShowSource (fps=24.013), output was in synch.

I have tried to give you a possible solution. It is up to you to try it.



Bye

I tried the 24.013 fps and cannot tell any difference from the 23.976 fps version. Both were in sync and produced the black lines at the top & bottom.

I used FLVSplitter, DirectShowSource and HCenc 1-pass. I used about a 30Mb clip & processing times were within a few seconds, but some of that depended on how quickly I hit a key in order to proceed.

I searched & tried to find out more about 24.013 fps but there’s not a lot of information out there. A few people mentioned it, and also 24.014, and if they should use it in a program. But I found nothing on how/when/why it is used. Why or when is it used?

[davexnet]

The frame count anomaly is based on the fact that the total amount of frames in the file is in question, although the runtime
is apparently agreed upon. For example with the test file, if you open it in virtualdub using the script, you see frame 24 is timed at
1.001 seconds - this is clearly 23.976 fps. I'm typing this from memory, but this method showed the file had 3108 frames,
while Gspot seemed to think the frame count was 3114.

[Gavino]

For example with the test file, if you open it in virtualdub using the script, you see frame 24 is timed at 1.001 seconds - this is clearly 23.976 fps.

It's hardly surprising you get 23.976 fps when your script calls DirectShowSource with fps=23.976!

However, I agree 23.976 is the correct rate.
If you open the file with FFVideoSource(..., timecodes=...), the timecodes file produced is exactly consistent with a rate of 23.976, except for the oddity that the first frame is timed at 83ms instead of 0. This is probably what is confusing everything and anything which tries to open the file. I believe GSpot's 24.013 is just spurious. Even FFVideoSource reports a rate of 23.8902, because it bases its calculation on the duration of the first frame, which is only accurate to the ms.

So davexnet's script with DirectShowSource(..., fps=23.976) works (audio in sync), as does
FFMPEGSource2(..., atrack=-1)
AssumeFPS(23.976)

[davexnet]

For example with the test file, if you open it in virtualdub using the script, you see frame 24 is timed at 1.001 seconds - this is clearly 23.976 fps.

It's hardly surprising you get 23.976 fps when your script calls DirectShowSource with fps=23.976!
)

Good point, however, in my mind it made sense at the time! Probably because it agreed with Mediainfo and produced an in-sync result.

I tried ffmpegsource2 as you mentioned and got a good result - Since it's not using dshow, you
don't get those black pixels at the top/bottom either.

LoadPlugin("J:\AVStoDVD_232_NoInstall\Lib\ffms2.dl l")
Import("F:\Documents and Settings\Dave Xnet\My Documents\Downloads\ffms2-2.13\ffms2-2.13\ffms2.avsi")

ffmpegsource2("F:\Documents and Settings\Dave Xnet\My Documents\Downloads\Hulu - Peacemakers.flv",atrack=-1,cachefile="K:\Temp\A2DFFV_" + String(Rand()) + ".cache")

ConvertToYV12
#Using DGPulldown/HCenc to upsize FPS
Lanczos4Resize(720,480)
assumefps(23.976)

[Mike99]

davexnet,

I’m somewhat confused on how all this fits together.

AVStoDVD has Preferences > AviSynth > AVS Source Filter > FFMpegSource.

But you are using ffmpegsource2. I presume I need to download an updated version, but where & how do you install it? I’m presuming the ffms2.dll is part of ffmpegsource2.

I’ve only worked with scripts once before, and that was about a year or two ago. I understand what they are but am not real familiar on how or where to use them. I do search for information, but I’m still lost on where to use your scripts such as LoadPlugin, Import and ffmpegsource2. Any help would be greatly appreciated.

Thanks in advance,
Mike

[davexnet]

Mike, I'm in AVStoDVD, edit title, Avisynth tab. All I did was delete what was there and copy/paste my new script in.
It will use that script and not revert back to it's original as long as you don't change anythng further in the edit title tabs.

You probably don't even need to do anything so drastic. In the preferences, select ffmpeg source. Start the project as new,
and when you set everything up, last thing to do, go to the Avisynth tab and add
assumefps(23.976) For example,
before the audiodub line, add video=video.assumefps=(23.976)

Did you see the other thread where I inquired about the black border? You'll find some more info in there.
https://forum.videohelp.com/threads/328121-FLVSplitter-FFDshow-FLV-resolution-anomaly

[Gavino]

But you are using ffmpegsource2. I presume I need to download an updated version, but where & how do you install it? I’m presuming the ffms2.dll is part of ffmpegsource2.

ffms2.dll contains the functions ffvideosource and ffaudiosource.
ffms2.avsi contains the function ffmpegsource2, which calls the other two.

If you install both files in your Avisynth installation plugins folder, they will be loaded automatically without need for LoadPlugin or Import.

[Mike99]

Mike, I'm in AVStoDVD, edit title, Avisynth tab. All I did was delete what was there and copy/paste my new script in.
It will use that script and not revert back to it's original as long as you don't change anythng further in the edit title tabs.

You probably don't even need to do anything so drastic. In the preferences, select ffmpeg source. Start the project as new,
and when you set everything up, last thing to do, go to the Avisynth tab and add
assumefps(23.976) For example,
before the audiodub line, add video=video.assumefps=(23.976)

Did you see the other thread where I inquired about the black border? You'll find some more info in there.
https://forum.videohelp.com/threads/328121-FLVSplitter-FFDshow-FLV-resolution-anomaly

I tried the above but got a couple of error messages. First message said there was a script error syntax error & then then second one said the process was aborted. Here is my script. I also tried adding blanks before & after the first equal sign so it had the same format as the other lines but it made no difference.

I did see the other thread & will have to go through it a couple times & digest the information.

LoadPlugin("C:\Program Files\AVStoDVD\Lib\ffms2.dll")

Audio = FFAudioSource("C:\Documents and Settings\Mike\My Documents\Hulu - Peacemakers Peacemakers-Pilot - Watch the full episode now..flv", track=-1, cachefile="C:\DOCUME~1\Mike\LOCALS~1\Temp\A2DFFA_" + String(Rand()) + ".cache")
Video = FFVideoSource("C:\Documents and Settings\Mike\My Documents\Hulu - Peacemakers Peacemakers-Pilot - Watch the full episode now..flv", track=-1, cachefile="C:\DOCUME~1\Mike\LOCALS~1\Temp\A2DFFV_" + String(Rand()) + ".cache", seekmode=0)

Video = Video.ConvertToYV12
#Using DGPulldown/HCenc to upsize FPS
Video = Video.Lanczos4Resize(720,480)

video=video.assumefps=(23.976)
AudioDub(Video, Audio)

[Gavino]

video=video.assumefps=(23.976)

There was a typo in davexnet's post. It should read

video=video.assumefps(23.976)

[Mike99]

But you are using ffmpegsource2. I presume I need to download an updated version, but where & how do you install it? I’m presuming the ffms2.dll is part of ffmpegsource2.

ffms2.dll contains the functions ffvideosource and ffaudiosource.
ffms2.avsi contains the function ffmpegsource2, which calls the other two.

If you install both files in your Avisynth installation plugins folder, they will be loaded automatically without need for LoadPlugin or Import.

It looks like you made a posting while I was typing mine. According to my script it appears ffms2.dll and ffvideosource and ffaudiosource are all there. It seems that selecting FFMpegSource in Preferences brings them into the script.

[djkilla]

MrC or anyone else:

If I'm using FFDShow with AVStoDVD, should I change AVS Source Filter (Preferences, AVISynth tab) to DirectShowSource or leave it at the default A2DSource?

If I'm using FFDShow with AVStoDVD and converting XVID (AVI) with AC3 audio to DVD, do I need to change FFDShow Mixer to 3/0/2+LFE or will AVStoDVD do this automatically?

[davexnet]

See this thread re: the mixer settings:
https://forum.videohelp.com/threads/318385-ffdshow-settings-and-conflicts?p=1974150#post1974150

In general, use A2dsource.The usage of directshowsource mentioned above is for a special case related to a problematic FLV source file.

[Mike99]

The corrected script worked for one test clip of Peacemakers using either HCenc or QuEnc.

However on another Peacemakers test clip if I use HCenc, it’s info screen states “Audio track contains no audio frames”. If I use QuEnc, I get an AviSynth error message stating the same thing.

I also tried a test clip of Fringe using HCenc & get the same message. The AviSynth is also slightly different. I tried it with the “video=video.assumefps(23.976)” inserted before either of the audio statements but it made no difference.

LoadPlugin("C:\Program Files\AVStoDVD\Lib\ffms2.dll")

Audio = FFAudioSource("C:\Documents and Settings\Mike\My Documents\Hulu - Fringe Amber 31422 - Watch the full episode now..flv", track=-1, cachefile="C:\DOCUME~1\Mike\LOCALS~1\Temp\A2DFFA_" + String(Rand()) + ".cache")
Video = FFVideoSource("C:\Documents and Settings\Mike\My Documents\Hulu - Fringe Amber 31422 - Watch the full episode now..flv", track=-1, cachefile="C:\DOCUME~1\Mike\LOCALS~1\Temp\A2DFFV_" + String(Rand()) + ".cache", seekmode=0)

Video = Video.ConvertToYV12
#Using DGPulldown/HCenc to upsize FPS
Video = Video.Lanczos4Resize(720,480)

Audio = Audio.SSRC(48000)

AudioDub(Video, Audio)

[Mike99]

I just tried the full length Peacemakers file and inserted “video=video.assumefps(23.976)”.

The first time using HCenc I got an info message “Incompatible input, try …..” I could not quite get the whole message because it disappeared but it had to do with YV12.

The next two attempts I got the audio track has no audio frames message.
I then tried QuEnc and got the no audio frames message.

I went back & tried the first Peacemakers test clip that used to work, but it does not work anymore. It now comes up with the no audio frames message.

I even closed & restarted AVStoDVD and keep getting “audio track has no audio frames”.


UPDATE -
I've been looking & double checking things. I tried the first test clip & it now works again. Unfortunately the second clip still does not.


davexnet -

The other thread that you started mentions the first frame is not the correct time length which was throwing off the fps readings in some programs. Could this be creating the no audio frames message I keep getting?

[Gavino]

I suspect the "no audio frames" problem is because you are re-using the same index (cache) file for different sources. Try just omitting the 'cachefile' argument, leaving it at its default.

Audio = FFAudioSource("C:\Documents and Settings\Mike\My Documents\Hulu - Fringe Amber 31422 - Watch the full episode now..flv", track=-1)
Video = FFVideoSource("C:\Documents and Settings\Mike\My Documents\Hulu - Fringe Amber 31422 - Watch the full episode now..flv", track=-1, seekmode=0)
...