Reversed video from AviSynth is "blocky".

[sambat]

Loading the script in VirtualDub, the video shows intermittent blockiness as it's stepped through.
Turning off direct decoding in Preferences and allowing ffdshow to take over makes no difference.
When the output is saved using the installed codecs - or an external codec - the same intermittent pattern shows.
Neither DVD nor AVI clips behave this way.

Feeding the script to MPC, same.
In either scenario, when the feed is not reversed, there is no problem.
Is it because it's a high resolution file or fast action or both;I tried resizing in the script but that made no difference.

AVCSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD_Track00.dga").reverse()

General
Unique ID : 211134710336940616268497285024418626123 (0x9ED710EB85049A65BD467FE06A47FA4B)
Complete name : E:\Office clip\_mkv\GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv
Format : Matroska
Format version : Version 1
File size : 53.2 MiB
Duration : 2mn 14s
Overall bit rate : 3 327 Kbps
Encoded date : UTC 2009-11-30 19:41:06
Writing application : mkvmerge v2.6.0 ('Kelly watch the Stars') built on Mar 24 2009 15:23:17
Writing library : libebml v0.7.7 + libmatroska v0.8.1
TITLE : G.I. Joe - The Rise Of Cobra Trailer
TITLE/Url : http://www.gijoemovie.com
COPYRIGHT : 2009 by PARAMOUNT PICTURES. All rights reserved. G.I. JOE and related characters are trademarks of Hasbro and are used with permission. 2009 Hasbro. All Rights Reserved.
COMMENT : Encoded in DivX Plus HD for the DivX Plus Web Player

Video
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L4.0
Format settings, CABAC : Yes
Format settings, ReFrames : 12 frames
Codec ID : V_MPEG4/ISO/AVC
Duration : 2mn 14s
Nominal bit rate : 3 200 Kbps
Width : 1 280 pixels
Height : 528 pixels
Display aspect ratio : 2.40:1
Frame rate : 23.976 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.197
Title : Main title
Writing library : x264 core 79 r1342 e8501ef
Encoding settings : cabac=1 / ref=12 / deblock=1:-1:-1 / analyse=0x3:0x133 / me=umh / subme=10 / psy=1 / psy_rd=1.0:0.2 / mixed_ref=1 / me_range=24 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / chroma_qp_offset=-3 / threads=8 / nr=0 / decimate=1 / mbaff=0 / constrained_intra=0 / bframes=3 / b_pyramid=0 / b_adapt=2 / b_bias=0 / direct=3 / wpredb=1 / wpredp=0 / keyint=95 / keyint_min=25 / scenecut=40 / rc_lookahead=60 / rc=2pass / mbtree=1 / bitrate=3200 / ratetol=1.0 / qcomp=0.60 / qpmin=10 / qpmax=51 / qpstep=4 / cplxblur=20.0 / qblur=0.5 / vbv_maxrate=20000 / vbv_bufsize=25000 / ip_ratio=1.40 / aq=1:1.00
Language : English
Default : Yes
Forced : No

Audio
ID : 2
Format : AAC
Format/Info : Advanced Audio Codec
Format profile : LC
Codec ID : A_AAC
Duration : 2mn 14s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 KHz
Compression mode : Lossy
Title : English 2ch Stereo
Language : English
Default : Yes
Forced : No

Text #1
ID : 3
Format : UTF-8
Codec ID : S_TEXT/UTF8
Codec ID/Info : UTF-8 Plain Text
Title : English
Language : English
Default : No
Forced : No

Text #2
ID : 4
Format : UTF-8
Codec ID : S_TEXT/UTF8
Codec ID/Info : UTF-8 Plain Text
Title : French
Language : French
Default : No
Forced : No

Text #3
ID : 5
Format : UTF-8
Codec ID : S_TEXT/UTF8
Codec ID/Info : UTF-8 Plain Text
Title : German
Language : German
Default : No
Forced : No

Text #4
ID : 6
Format : UTF-8
Codec ID : S_TEXT/UTF8
Codec ID/Info : UTF-8 Plain Text
Title : Portuguese
Language : Portuguese
Default : No
Forced : Yes

Text #5
ID : 7
Format : UTF-8
Codec ID : S_TEXT/UTF8
Codec ID/Info : UTF-8 Plain Text
Title : Spanish
Language : Spanish
Default : No
Forced : No

[Selur]

try:

AVCSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD_Track00.dga")
RequestLinear(rlim=50,clim=50) # comes with TIVTC
reverse()

if that doesn't help try ffmpegsource2 or LWLibavVideoSource instead of AVCSource.

[jagabo]

When frames are requested out of order many decoders will have problems with codecs that encode frames out of order. They get "confused" and start making gross errors like that. If you can't get the decoder working you'll have to resort to reading the file linearly and saving to a lossless intermediate. Then reverse that.

[sambat]

Thank you.
AVCSource and ffmpegsource2 both give the same error in VirtualDub.
(MPC displays the same error when opening the script).
Also, can't locate the plugin LWLibavVideoSource

try:

AVCSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD_Track00.dga")
RequestLinear(rlim=50,clim=50) # comes with TIVTC
reverse()

if that doesn't help try ffmpegsource2 or LWLibavVideoSource instead of AVCSource.

[sambat]

When frames are requested out of order many decoders will have problems with codecs that encode frames out of order. They get "confused" and start making gross errors like that. If you can't get the decoder working you'll have to resort to reading the file linearly and saving to a lossless intermediate. Then reverse that.

Which scenario is more likely: AviSynth is presenting the errors to VirtualDub or VirtualDub is creating the problem.

[poisondeathray]

Try

FFVideoSource("video.mkv", threads=1, seekmode=0)
Reverse()

[jagabo]

Try

FFVideoSource("video.mkv", threads=1, seekmode=0)
Reverse()

Yes, that might work. But it will be incredibly slow if the video is long.

[poisondeathray]

Try

FFVideoSource("video.mkv", threads=1, seekmode=0)
Reverse()

Yes, that might work. But it will be incredibly slow if the video is long.

It's 2min14sec if mediainfo is correct (post #1)

Your lossless I-frame encode suggestion should definitely work (and I would use I-frame, not long GOP lossless for this)

[Selur]

Sad, I thought that AVCSource would use the same frame cache DGSource with Avisynth MT uses, but that doesn't seem to be the case, seekmode=0, with FFVideoSource like poisondeathray suggested could also help.

LWLibavVideoSource is part of the LSMASHSource.dll -> https://forum.doom9.org/showthread.php?t=167435

[sambat]

Try

FFVideoSource("video.mkv", threads=1, seekmode=0)
Reverse()

Many thanks;that worked with no blocking.
It took 06:49 to process out of VirtualDub, without resizing, using the external mp4 codec.

[poisondeathray]

How long did it take normally on your setup ? (encoding without threads and seekmode, but with reverse, resulting in the "blocky" result ?)

As jagabo alluded to, the longer the clip duration, it becomes almost exponentially slower. There is a point at which encoding to a lossless I frame intermediate will be faster

[sambat]

It took almost 14 minutes with the original script as opposed to 6:42 with the last script.

[poisondeathray]

It took almost 14 minutes with the original script as opposed to 6:42 with the last script.

???? well I'm completely confused by that . Did you mean AVCSource() for the "original script" , or FFMpegSource2() or something else

Are you sure the encoding settings, all other conditions etc.., audio encoding, were the same ?

[sambat]

The only difference is the scripts.

AVCSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD_Track00.dga").reverse()



FFVideoSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv", threads=1, seekmode=0).Reverse()

[Selur]

*gig* there are various reasons why AVCSource isn't recommend,...

[poisondeathray]

Don't worry , it was only for curiosity

What I meant to compare was

FFVideoSource().Reverse()

VS.

FFVideoSource(threads=1, seekmode=0).Reverse()


Because AVCSource and FFVideoSource are using different decoders

[Selur]

Because AVCSource and FFVideoSource are using different decoders

not that different, AVCSource is based on an older libav version

btw. threads=1 is probably not necessary nowadays,..

[poisondeathray]

Because AVCSource and FFVideoSource are using different decoders

not that different, AVCSource is based on an older libav version

This is true, In my experience the older libav version neuron2 uses is much slower, and (I think single threaded) . Even years ago, when people used AVCSource vs. DirectShowSource with ffdshow (libav) , DirectShowSource was much faster

What else can account for sambat's observed ~2x speed difference (even when ffvideosource is using threads=1, seekmode=0 !!!!!!!)

btw. threads=1 is probably not necessary nowadays,..

Most of the time I would agree, but there are cases where you can demonstrate that this still makes a difference .

[Selur]

Most of the time I would agree, but there are cases where you can demonstrate that this still makes a difference .

haven't encountered such a case for quite some time (but I'm mainly using ffmpegsource for MPEG-4 ASP/AVC content)
-> got an example case?

[poisondeathray]

Most of the time I would agree, but there are cases where you can demonstrate that this still makes a difference .

haven't encountered such a case for quite some time (but I'm mainly using ffmpegsource for MPEG-4 ASP/AVC content)
-> got an example case?

some interlaced avc transport streams (and even with threads=1, and other tricks like remuxing to mkv first, can still pose problems)

For MPEG4-ASP, it usually doesn't matter in my experience, but someone was having problems the other day (it wasn't clear to me if the fix was from using threads=1, or if using the vanilla version of ffms2.dll vs. cplugin; I'm still unclear as to why he got the green frame with ffms2, I couldn't reproduce it here)
https://forum.videohelp.com/threads/354878-Convert-AVI-to-MPEG4-AVC-always-2-lost-frames

[Selur]

interlaced content seems to be nearly always broken when ffmpegsource is used (no matter the thread count) and the other thread looks more like a version problem :P

[poisondeathray]

interlaced content seems to be nearly always broken when ffmpegsource is used (no matter the thread count)

This is true but there are cases where it helps. There were examples in the FFMS2 thread.

There are also cases where it helps with other types of files, a few posted on this board. I don't have them all archived, but you should do a search if you're still interested. Many of them had to do with seek corruption and certain ffms2 builds

This one was actually from a thread at videohelp, I'm certain there are a few more, because I can remember them.
http://code.google.com/p/ffmpegsource/issues/detail?id=74

It was fixed in one of the newer revisions, but broken, then fixed again. Not sure what the current status is.

PS. whenever there is a problem with FFMS2, first thing I do is try seekmode=0, threads=1.

and the other thread looks more like a version problem :P

I couldn't reproduce it with or without Cplugin..so I don't know what caused or fixed it...

[poisondeathray]

And for interlaced AVC transport streams, DGNVTools is preferred, but alternative workaround is FFMBC lossless intermediate.

FFMBC still handles interaced transport streams better than FFMPEG ; there was a FFMBC based FFMS2 compiled, but it doesn't have the same "magic" as FFMBC binary

[jagabo]

I've seen that "insanity detected" many times with ffmpegsource. I didn't know about the threads=1 fix. I just switched to another decoder. Just today I had that problem. I tried threads=1 and it worked. Then I updated to a newer build. And no longer needed threads=1.

[poisondeathray]

I've seen that "insanity detected" many times with ffmpegsource. I didn't know about the threads=1 fix. I just switched to another decoder. Just today I had that problem. I tried threads=1 and it worked. Then I updated to a newer build. And no longer needed threads=1.

The "insanity detected" is a (fairly?) common example, but another one is seek corruption (macroblocking) which is sometimes fixed with threads =1

It's hard to keep track of which version works, doesn't work. They keep on breaking/fixing things.

[sambat]

Don't worry , it was only for curiosity

What I meant to compare was

FFVideoSource().Reverse()

VS.

FFVideoSource(threads=1, seekmode=0).Reverse()


Because AVCSource and FFVideoSource are using different decoders

03:55 and 06:13 respectively.

[sambat]

Continuing on, I can't get audio unless I load it separately in VirtualDub.
Converted the AAC to AC3, to no avail.
There is no error message, but VirtualDub file information does not show audio.

a=FFVideoSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv", threads=1, seekmode=0).LanczosResize(640,264)
b=FFVideoSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv", threads=1, seekmode=0).LanczosResize(640,264).Reverse()
vid=StackHorizontal(a,b)
#audio=ffaudiosource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv")
audio=nicac3source("GIJoeTheRiseOfCobraTrailer-DivXPlusHD_Track01.ac3")
audiodub(vid, audio)
return(vid)

[poisondeathray]

03:55 and 06:13 respectively.

Thanks, that's more in line with expectations

Continuing on, I can't get audio unless I load it separately in VirtualDub.

What about a simple script without reverse or filters? check info() , what does it say ?

vid=FFVideoSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv", threads=1, seekmode=0)
aud=FFAudioSource("GIJoeTheRiseOfCobraTrailer-DivXPlusHD.mkv")
audiodub(vid,aud)
info()

[sambat]

Audio OK.

[poisondeathray]

And vdub shows audio with that script loaded (file=>file information) ?

If it does, add back the resize and reverse to vid , if audio is still ok, then add the 2nd video and stackhorizontal