870717 - converting TS to Xvid

[hamidi2]

u all did favors to me and i really appreciate ur assistance
yeah, yadif is a very good deinterlacing method for the clips, thx
i think the problems r solved now :*
now i know where to post my future problems in case they happen
cer10ly without ur help i couldn't succeed
thk u very much
regards

[hamidi2]

Oops!
tsmuxer doesn't provide a synced output always

[poisondeathray]

TsMuxeR can only repair small errors.

If you have big errors in your stream you will have to process in segments manually and/or cut bad frames out

[hamidi2]

i wonder. why this out of sync must essentially occur?
why a/v r synced when the .ts file is played in vlc? what can it do that tsmuxer can't? tsmuxer says that for some .ts files more than one resync must be done. how does it recognize this out of syncs and how can it realize that how much the audio must be time shifted? if it's a value which may be extracted from the source, why does it mistake some times?
it's weird! when i play the same .ts file with vlc or play it in the box (satellite receiver, dreambox), everything is ok and there's no out of sync. how do vlc and the box don't mistake and realize the exact values that the software designed for this purpose, tsmuxer, even can't?...
might this be because i've combined the 1GB files together to make a bigger standalone file for softwares to be able to handle it?

[poisondeathray]

It might be the error is introduced when you join the files. How are you joining them? before or after conversion? Are you editing commercials -because this may be the step that introduced the desync. Try converting first and editing later instead

When you are playing back, the software often ignores or speeds up/slows down on bad frames and has error recovery. VLC is great for playing back damaged streams

When you are transcoding, the encoder usually drops a frame - either audio or video or both - this will cause a desync problem.

You can try indexing using the DGMPGdec method, but if the error occurs between timestamps, it will still be out of sync. I don't think it will work, because avidemux indexes your file as well

Another option which worked on a h264/aac .ts with 100's (!) of errors (amazing that it worked because many other approaches didn't work) was to use avisynth using the directshow input driver (placed in the avisynth plugins directory.)
http://www.virtualdub.org/beta/DShowInputDriver-0.5.zip

In your case, you have to enable MPEG2 in the ffdshow configuration to work (other MPEG2 decoders don't work for MPEG2 .ts very well). Open the .ts using the drop down dialog to select the directshow input driver specifically. Select video=>fast recompress then video=>compression , xvid and configure your settings. Select audio full processing mode, then audio=>compression (e.g. LAME MP3 and configure your settings). File=>save as avi

[hamidi2]

i joined them as they were not video files. like copy /b command, i used total commander to join them ignoring their types. as i had experienced b4 for .vob files it's not important how they get splitted. they're indeed a big file splitted in smaller parts for DVD players which not support long files to be able to play them back. i thought that .ts files must be similar to .vob files. i tried joining them and foind that my conclusion was right because the resulting file might be played back by vlc with no problem.
my problem is b4 editing them. even b4 converting them. avidemun is not able to play them back synced. this is the source of the problem. b4 everything the tools may be able to recognize any out of sync of the source file to be essentially able to play it back correctly. then we may think about conversion.
again, i tried to play the 26GB file back by the vlc to make sure that there's no out of sync in any part of the .ts file and found that it works well.
is there any tools like vlc to be able to recover the input stream from any errors and be able to play it back correctly? b4 manipulation on any stream it's necessary to recognize it completely. avidemun, i think is not able to figure out all damages.
as i said, my problem is b4 transcoding.
i will examine DGMPGdec though.
a thing i can't figure out is why the source stream must be essentially damaged?!
the method u introduced about avisynth sound a bit more professional that what i can figure simply out. i will do my best about it and will give u my feedback.
i can't say how much i'm favored and am grateful about ur assistance and the time u spent about my problem.
god bless u, and thx so much.
i will hope to be able to solve my problem by ur assistances

[poisondeathray]

In my experience, joining videos through DOS commands never works. It usually doesn't preserve header information or data specific to the container

I would try avidemux, and use the file=>append to join clips.

If that doesn't work try DGMPGdec (it uses avisynth so you will have to do some reading - start with the included quickstart manual)

Last resort would be the vdub directshow input driver method - again I'm very impressed that this worked on a clip with more almost errors than "regular" sections, and all other methods failed

Good luck

[hamidi2]

i downloaded avisynth and installed it. but didn't found any .exe file in the program files. i didn't know how to use it.
then i decided to download the dshowinputdriver. since it's from virtualdun site, i thought that it must a plugin for the tools. i copied it into plugins32 folder of virtualdub because of it's extension. i then dropped the .ts file into virtualdub, but it said that it didn't recognize this type of stream. then i tried to open it and selected the directshow files from the combobox, and wow! now it got open. the 1st thing i did anxiously was to try converting a little part of the .ts file. and fortunately i found it worked!
is this the best way to convert the stream? since there's no out of sync, may i use virtualdub as the standalone converter and editor without having to use any other additional utilities? what is the difference between avidemux output and the virtualdub output?
is this the final solution?

[hamidi2]

yeah, i know that this method doesn't preserve header information but i think for .ts and .vob files the header information is located for the whole movie only in the 1st part with the extensions of .ts and .vob, and not repeated in .001, .002, etc files. they're just saved as another file to prevent large files problems which may occur in some dvd players. it's not my knowledge, i just feel so.

[poisondeathray]

Yes you have to "activate" the plugin by using the drop down box then selecting the .ts

There is no "best" way. This is just 1 option that seems to work with damaged .ts files (it even records the errors and glitches as if it were played in vlc or a media player)

I would recommend using "append" in vdub, instead of using DOS commands.

Yes you can use vdub only as the converter and editor, but you may get desynch using this method (you may introduce more error). I would give it a try though, since you've come this far

The "safe way" would be to convert it to an uncompressed avi, edit that, then convert to your final xvid/mp3/avi format all in vdub

The difference in output between avidemux and vdub depends on what codecs/settings you use

Good luck

[hamidi2]

i use xvid codec with the same settings for both, so i would probably get the same output. right?
it's unpractical to convert the source steam into an uncompressed avi! the source is a 26GB file, man!!!
i will examine more files and will try more parts and will give the result in case of desync.
thx

[jagabo]

If VirtualDub's DirectShow source filter worked then AviSynth's DirectShowSource() filter should work too. Then you could do all your processing in AviSynth. If you do any filtering in VirtualDub you will lose some color accuracy when VirtualDub converts to RGB for processing.

[poisondeathray]

If VirtualDub's DirectShow source filter worked then AviSynth's DirectShowSource() filter should work too. Then you could do all your processing in AviSynth. If you do any filtering in VirtualDub you will lose some color accuracy when VirtualDub converts to RGB for processing.

Good idea jagabo, I would try that first.

It's funny - that's what I usually do. Logic says if it plays ok in a directshow player, it should work with DirectShowSource() right?

But there are some streams that have so many errors that DirectShowSource() doesn't work (either the video or audio gets dropped in several places), and yet it plays ok in a media player.

Here is an example (it's a badly damaged h264 .ts, so kind of off topic here...). I tried every trick in the book, different .avs scripts, ffmpegsource(), avcsource(), different filters, different decoders, different splitters, changing containers, indexing with different programs, processing audio & video separately & together, dozens of different programs, repairing the .ts, and the only way that worked was the vdub input driver (I don't know why it should be different than DSS() ). It even copied the errors precisely in the exact frames/timespots with corresponding audio glitches. Note: I even tried copying the audio with a hardware recorder as it was playing in the software player, but the running time ended up different! Impossible! Credit to "canTsTop" who figured this one out.

http://www.mediafire.com/file/azxjyatwffn/with_errors_LTV.ts

[jagabo]

Nice video! LOL I've added it to my collection of troublesome videos.

It plays (aside from the obvious problems) without A/V sync errors in all the players I've tried. AviSynth's DirectShowSource() resulted in A/V sync errors even with convertfps=true and fps=25. DgAvcIndex was able to index it but VirtualDub (AvcSource()) crashed while encoding. The partial encoding had sync errors.

I can't get VirtualDub to open the TS file at all with the Directshow source plugin (it locks up). What TS splitter are you using? What h.264 decoder? I'm using recent release of CoreAVC Pro and the version of Haali Media Spolitter that came with it.

[poisondeathray]

fun clip

vdub 1.8.5
directshow input driver 0.5
haali media splitter (the version says 1.8.122.18 in the icon bar)
ffdshow for h.264 (coreavc pro 1.7 and 1.8 don't work)

I tried every combination I could think of - even using graphstudio/graphedit .grf files and exotic decoders like elecard, arcsoft, mainconcept.

There is no logical reason why this should work, but it does for some reason. I'd love to hear if you have another solution, especially through avisynth

Here is the "successfully" converted file in sync
http://www.megaupload.com/?d=QWW0QK1U

(I apologize for the thread hijack )

[jagabo]

I have the same version of Haali. I don't usually use ffdshow's h.264 decoder because it doesn't work well with my Hauppauge HD PVR's TS files and MPCHC. But I enabled it for a test with VirtualDub's DS source filter. It worked and I was able to encode with the audio and video in sync. AviSynth with Haali and ffdshow still gives A/V sync errors. I'll play around with this some more tomorrow. Thanks for the help.

[hamidi2]

hi
please don't confuse me with ur profession
it sounds lots of tools a video handler must struggle with!
but i just come again with another problem. like always i thought my problem was resolved at last, this time i've a new problem.
the resulting converted video by vdub is not in width as the same as the original one. vlc doesn't give me the properties. vdub and avidemux say that it's a 544x576 video, while when it's played in vlc width is visually more than height.
another problem is that i don't know which option in vlc is similar to avidemux's yadif method for deinterlacing.
avidemux would be the best solution if it was synced
i also have no idea how to use avisynth
waiting anxiously to hear from u
i'm also going to do the same thing i did with my video with the one u put in mediafire and give the result. this will help u figure out what mistakes i might have.
thx

[jagabo]

the resulting converted video by vdub is not in width as the same as the original one. vlc doesn't give me the properties. vdub and avidemux say that it's a 544x576 video, while when it's played in vlc width is visually more than height.

MPEG files have a Display Aspect Ratio (DAR) flag. The DAR flag tells a player what final aspect ratio should be seen on screen, regardless of the frame dimensions with which the video is encoded. Your video is encoded with a frame size of 544x576 and has the DAR flag set to 4:3. VLC is automatically scaling the image to match the DAR flag.

VirtualDub's default behaviour is to map the source's pixels one to one with pixels on the monitor. This lets you see each pixel without any distortions or artifacts created by resizing. If you want to see the frames with the correct DAR, right click on a pane and select the DAR. Note that VirtualDub uses quick dirty resizing filter for display (equivalent to Nearest Neighbor in its Resize filter) so it won't look great.

When encoding to Xvid AVI you can retain the frame size and set Xvid's DAR flag or you can resize the frame yourself and use square pixels. Not all players will respect the DAR flag so resizing the video is more common. After deinterlacing you can resize your 544x576 video to 768x576, 544x408, 512x384, 448x336, or some other 4:3 ratio.

another problem is that i don't know which option in vlc is similar to avidemux's yadif method for deinterlacing.

X is closest to Yadif in mode 0, Linear in mode 1. The former has better resolution, the latter better motion.

i also have no idea how to use avisynth

AviSynth is the most flexible and powerful video processing tool. It is controlled via text based script files. You can create the script files with any text editor (like Notepad) or via some more automated tools like FitCD or AVSP. After creating an AVS script you open it with your editor/encoder just as if it was a video file. AviSynth does all the work and passes frames one-by-one to the program.

A full discussion of AviSynth would take months to write. But a simple script to do what you want might look like:

Load_Stdcall_plugin("C:\Program Files\AviSynth 2.5\plugins\yadif.dll")
DirectShowSource("Filename.ts")
Yadif(mode=0, order=1)
LanczosResize(512, 384)

Note that Yadif is not included in the AviSynth package. You must download and install it separately. Most filters will load automatically if placed in AviSynth's plugins folder, but Yadif requires manual loading (as depicted in the above script).

[hamidi2]

i think i must pass some courses to earn the same experiences and something like avisynth is a too professional tools for me to get involved. i just need a simple tools for cutting and converting. that's all.
i didn't see anything in the context menu of vdub named DAR. i just found 4:3 TV which made it the same resolution i wanted. u meant it will keep the DAR flag in the destination converted avi file? if so, i've not to worry about resolution? but i think as far as i want to be able to play it properly by TV and also vlc, it's better to make the destination avi file as i wish it to be. don't u think so?
and about the deinterlacing, i didn't find the X option in the deinterlace filter of vdub. it just includes Blend fields together (best), Duplicate field 1, ... to Fold side-by-side fields together.
what i want from deinterlacing is to exactly be shows with the same quality in vlc as it does (the original one) in the TV. i don't know how the receiver box plays it back in the TV. i know that no interlace lines is seen in the TV and i like to be able to play the same quality and same exact thing i do now after conversion too both on TV and on computer. with whatever player (even if not able to deinterlace at play back like vlc) i use.
sorry, i know that i could not explain so good. i hope u may understand completely
thx

[jagabo]

i didn't see anything in the context menu of vdub named DAR.

i think as far as i want to be able to play it properly by TV and also vlc, it's better to make the destination avi file as i wish it to be. don't u think so?

It's safest to resize to a 4:3 frame size.

and about the deinterlacing, i didn't find the X option in the deinterlace filter of vdub.

You asked about VLC, not VirtualDub.

it just includes Blend fields together (best), Duplicate field 1, ... to Fold side-by-side fields together.

Use Discard Field 1 or 2, then resize what's left. Or you could try one of the third party deinterlace filters like Deinterlace - Smooth, or Muksun.

what i want from deinterlacing is to exactly be shows with the same quality in vlc as it does (the original one) in the TV.

You can't really do this without doubling the frame rate. Try VirtualDub's Bob Doubler. But the results probably won't play on a Divx certified DVD players.

[hamidi2]

thanx 4 ur so prompt and complete response
i didn't mean frame rate by the quality. i only concentrated on the deinterlacing. i meant i liked to see no deinterlacing and see exactly the same thing i see in the tv by playing back the original file.
what's the difference between discard field 1 and 2? i may select whatever i like? and there'll be no difference in the output result?

[hamidi2]

as the word "discard" indicates some data gets lost?!
i don't like it to be resized to a 544x288 video and then get scaled again to 544x576 by doubling the lines!!!! this is exactly what i meant about quality! in such a case the resulting video will differ from the source one certainly and include half data of the original movie!
i like to save as much as data possible. i like to keep the same resolution. but just discard deinterlacing.

[jagabo]

i didn't mean frame rate by the quality. i only concentrated on the deinterlacing. i meant i liked to see no deinterlacing and see exactly the same thing i see in the tv by playing back the original file.

You don't understand. When you watch 25 fps interlaced video on TV you see 50 fields per second. You see one field at a time, never an entire frame. This can only be duplicated on a computer monitor by bob'ing the video. You can do this by encoding interlaced and having the player bob as it plays, or by bob'ing before encoding and creating a 50 frame per second AVI file.

If you want 25 progressive frame per second you will lose temporal resolution (motion will not be as smooth) and at least some spacial resolution:

Discard Field and Resize in VirtualDub is equivalent to half a Bob. You lose half the spacial resolution and half the temporal resolution but it's fast and easy. You will get some aliasing artifacts on sharp, near horizontal edges.

Smart Deinterlace by Donald Graft tries to keep as much of the second field as possible. You will lose half the temporal resolution but less than half the spacial resolution (parts of the frame that aren't moving aren't deinterlaced). You will get aliasing artifacts on moving, sharp, near horozontal edges. This is one of the better deinterlacers in VirtualDub.

Deinterlace 'Muksun' is, I believe, another motion adaptive smart deinterlacer. I haven't used it much but it appears to be similar to Donald Graft's.

[hamidi2]

thank u for such a complete info about deinterlacing. i think now i may figure out the nature of deinterlacing. if i may stop time at the end of each vertical sweap of a tv set, i see that only alternate lines r shown. first even lines, and in the next sweap which occurs in the next .02 seconds old lines r shown. this enables less data to be transfered with the same quality for eyes. this enables bandwidths to be used by half. did i get it correctly?
the reason i never see the artifacts and sharp edges on a tv is that they never appear together on screen. even lines and odd lines r drawn separately.
ok. if i got it correctly until now, i prefer it to resemble to what happens in a tv set. i don't know what's bob'ing. i think i've to think about a 50 fps avi because i don't want to lose any frame. then i think the video will be half in height, because i'm not going to produce a video with black lines. all data will be available in the output video and i've just to double the height during conversion. i think this is the only way to keep all data and have no lost. am i right? and if it is so, how can i do that?
thx

[hamidi2]

in the deinterlace filter of vdub, if i choose duplicate field 1, does it ignore field 2 and is it equal to just paying attention to only 25 frames in which only even lines have data and is it equal to discard field 2 and then doubling the height?

[hamidi2]

i think i have to select bob doubler filter in vdub. i did it with the default configuration, Bottom field first, and Bob as the deinterlacing method. is this exactly what u meant?
it generated the same resolution in output and 50 fps. but:

1. with this method i think i will get a double-sized file output but not double quality. so it seems that it's not worthy to bear with that. what's ur suggestion? converting to a lower frame rate avi file will certainly decreases the quality. so it's reasonable to admit lower fps in a way not to be sensible with eyes.

2. what would be the difference if i choosed top fields first? or used another deinterlacing method (ELA or Adaptive ELA)

thx

[hamidi2]

3. what does the Blend button do?

[jagabo]

2. what would be the difference if i choosed top fields first?

Each frame contains two pictures (fields). On an interlaced TV you see one field at a time. The field order determines which of the two fields you see first, which is seen second. If you use the wrong field order your video will jerk back and forth 25 times a second whenever there is motion.

or used another deinterlacing method (ELA or Adaptive ELA)

Those two settings try to reduce the aliasing at moving, sharp, nearly horizontal edges while bob'ing.

what does the Blend button do?

Are you talking about the regular deinterlace filter? It blurs the two fields together resulting in a fuzzy image. It looks ok with small motions but when motions get large it looks like double exposures.

[hamidi2]

how can i find out which one to choose? how can i find out the order of fields i mean?
and about the remainder no one has an idea?
thx

[jagabo]

how can i find out which one to choose? how can i find out the order of fields i mean?

Pick one, single step through the video with the right arrow key. If motion in the output window is smooth you picked the right one. If motion moves back-and-forth, you picked the wrong one. DV is bottom field first. MPEG, more often than not, is top field first.

1. with this method (bob doubler) i think i will get a double-sized file output but not double quality. so it seems that it's not worthy to bear with that. what's ur suggestion? converting to a lower frame rate avi file will certainly decreases the quality. so it's reasonable to admit lower fps in a way not to be sensible with eyes.

I'm not sure what you're asking. Doubling the frame rate may not double the file size but it will require a bigger file to retain quality. The resulting video will look very much like watching the original video on an analog CRT TV. I gave you several other options already.

Take a look at http://www.100fps.com for some deinterlacing examples.

Here are some examples I did a while back:
https://forum.videohelp.com/topic354397-30.html#1877202
I was mostly concentrating on the seams (black lines) between the panels of the truck. TempGaussMC_beta1 did the best job.