Motion problems with Canopus and TMPG

[Bizuser]

I bought a Canopus ADVC 100 to convert VHS tapes to dvd (I started out using my ATI AIW card and VDub, but gave up fighting audio synch problems).

1) I capture using premiere, saving to AVI.
2) I convert the Microsoft DV coded AVI to Canopus DV using the free Canopus utility, so I can use TMPG.
3) I use TMPG to convert the Canopus DV encoded AVI file to MPEG-2.
4) I then author/burn to DVD using TMPG's authoring program.

Results have been good - given that my videos are pretty old, I didn't notice any problem.

But this week I started using the same setup to capture some live TV shows from Cable. I've previously used my ATI AIW for this, but again, the audio synch problems have caused me to abandon this. I've also tried recording direct to MPEG2, but since I want to edit commercials, editing MPEG2 is always an issue.

The results have been 'good' - great image quality, perfect audio synch but ... 'motion' does not look too great in the final product. People moving, panning scenes, etc - don't look super-smooth.

I've learned that viewing the output on the PC is not a good guide, so I always burn to DVD and test on a set-top player. But if I look at the AVI files on the PC, they do appear to have very smooth movement, while the MPEG-2 files do not. So - could the poor motion effect be caused by the MPEG encoding?

Is it essential to use the highest 'motion search' setting in TMPG, and to use 2-pass VBR, to get top quality conversions?

Another question ... I have no particular desire to use Premiere for this purpose, and the fact that it outputs using the Microsoft DV codec is a pain, since I have to convert to the Canopus format before encoding in TMPG. Are there any freeware programs that can capture the Canopus input?

Is there a way to have Premiere do a 'fast save' of the AVI file if I do nothing more than cut out commercials? In my mind, if I simply cut out frames, I should not have to go through a several-hour rendering phase to save the output; cutting frames should be a special case that can bypass rendering? Is there a basic editing program that will do this?

Thanks!

[thecoalman]

1) I capture using premiere, saving to AVI.
2) I convert the Microsoft DV coded AVI to Canopus DV using the free Canopus utility, so I can use TMPG.

You should be able to eliminate step two by either configuring Premier to use the Canopus Codec, or use the capture app that came with the canopus or use premier to create your mpg. Seems odd that TMPG won't accept the file you have now, I don't use it so I'm not familiar with it's quarks.

Is there a way to have Premiere do a 'fast save' of the AVI file if I do nothing more than cut out commercials? In my mind, if I simply cut out frames, I should not have to go through a several-hour rendering phase to save the output; cutting frames should be a special case that can bypass rendering? Is there a basic editing program that will do this?

I think your going about it the wrong way. This is how I do it. I use Media Studio Pro which is the equivalent of Premier. I capture the footage to AVI, I put it in MSP and do my edits. Once my edits are complete I use MSP to create DVD compliant mpg's or whatever format I want. Put it in my authoring app and burn to disc, they are already compliant so they don't have to be rendered.

Premeire should do a "fast save" providing you just cut frames out and do not change any video settings, add transitions, filters etc. It's not exactly a fast save, I just ran a test on my comp with MSP. 1 minute 200mb AVI took 30 seconds on the same drive the source is on and 15 seconds on a different drive.

[Bizuser]

You should be able to eliminate step two by either configuring Premier to use the Canopus Codec, or use the capture app that came with the canopus or use premier to create your mpg. Seems odd that TMPG won't accept the file you have now, I don't use it so I'm not familiar with it's quarks.

My canopus did not come with any software whatsoever. Did yours? I seem to recall reading that other users commented that they too did not receive software. This seemed odd to me at the time!

Canopus's website has the codec convertor, but not a codec itself - as far as I know. This is where I got the convertor from: http://www.canopus.us/US/products/DV_file_converter/pm_dv_file_converter.asp
I cannot find an actual codec that will allow me to encode straight to Canopus format. The Microsoft DV format is the only option in Premiere. If you know of a codec I could install, that would be great.

As for using Premiere to create the mpeg directly, that is logical but, my version at least can't do it (premiere 6.0). It only has mpeg-1 capabilities. I guess later versions included a main concept mpeg2 encoder.

Premeire should do a "fast save" providing you just cut frames out and do not change any video settings, add transitions, filters etc. It's not exactly a fast save, I just ran a test on my comp with MSP. 1 minute 200mb AVI took 30 seconds on the same drive the source is on and 15 seconds on a different drive.

That's about the same as my results - about 18 minutes to process a 30 minute clip. That seems long when "all it should be doing" (ha ha) is dropping info from the file and possibly re-writing timing markers. I've solved this for now by doing my edits in tmpg; it has a pretty good interface that lets me drop commercials, etc (the 'source select' part of the advanced tab).

But my main concern/question is about the quality of 'motion' in the captured video. I captured a Seinfeld episode as a test, and in scenes where people are walking down the street, the movement looks a bit jerky. Using the absolute best quality settings (2-pass vbr, plus highest quality motion search) in tmpg makes it almost perfect, but this takes an age to process (1 hr 15 mins to process 15 minutes - 5:1 - on my P4 2.6 MHz machine.

[vhelp]

In my experience, the first thing to check for WHY there is audio
sync problems, is to be absolute cuirtain that that is no frame
drops. All you need basically, is just one frame drop to
throw off the audio. Doesn't really matter what capture app
you use either (in most cases) but that the first step to eliminating
that number one cause in audio sync issues.

The next cause is most likely how you are encoding your source. For
instance, it could be:

* IVTC methods
* Filters that alter Fields or Frames (ie, Resizing)
* Field Order (maybe)
* Your System limits; Codec limits; or Capture device's limits

First determine what the nature of the video source is. This is crucual
to successful transfer projects. For instance:

* TV show (ie, News; Sports; Soaps; Game shows; Live Concerts) these are
... all Interlace (every frame) = Interlace

* TV movie (ie, HBO; Cinemax etc) = Film (Telecined)

* VHS (commercial movie) = Film (Telecined)

* VHS (sporting event; news; soaps; games; live concerts, etc) = Interlace

Anything that is Interlace, requires that you setup your encoding params
using 29.970 fps only.

Anything that is Film (Telecined) requires you to know for sure that the
source program is Pure Film (Telecined) properly without any issues and
pefect 3:2 pulldown applied. Unfortunately, there are many tv programs that
are Film, but Telecined either: Badly or on Tainted on Purpose.
But, once the Film 's nature is determined (ie, Telecined properly) you can
safely apply an IVTC method to your encoding (to bring the the Telecined 29.970
fps capture, down to 23.976 fps, which equals Film)
Note, if you have a true Film source, but you have frame drops in your captured
source, your IVTC method will not work properly.., causing studders or other
problems during DVD play.


About Hardware Capture Devices..

Unfortunately, these devices do not offer any Built-in IVTC in their captures.
They simply only offer straight Captures to MPEG.. That's it. Therefore, you
cannot apply such IVTC processes to these, unless you re-encode these MPEG's.

That's about it,
-vhelp

[thecoalman]

Canopus codec, http://www.canopus.com/US/Products/DV_Codec/pm_dvcodec.asp BTW the link for this can be found in the tools menu. Lot's of goodies to the left.

Canopus capture instructions using supplied software http://www.canopus.com/us/support/tut_convert_analog.asp Maybe you have a older version without the software? I dunno that's just a guess. The canopus dv codec probably would have been installed with this app.

As for premier and not being able to do MPG2 that really suprises me and I doubt it, you probably only need a mpeg2 codec installed (possibly in the tools section, take a look). The amount of time it takes to create new file is mostly disc write time, time it then take the same file and and make a duplicate file you'll see what I mean.

[Bizuser]

In my experience, the first thing to check for WHY there is audio sync problems...

That's about it,
-vhelp

Very interesting information, but ... I'm not having audio synch problems at all now - that's why I bought the Canopus ADVC 100. <update> If you are passing this on because I said I originally had audio synch problems, that was due to trying to capture using vdub with my ATI AIW card. I had pretty good results using vdub and mjpeg compression (good compromise between quality and size), but I would get audio synch due to dropped frames AND - I believe - audio drift (lack of synch due to audio signal being processed by different hardware - sound card - rather than video capture card. Anyway, it worked pretty well, but that plus the sheer size of the files caused me to buy the canopus, which simply seems to work better.

My issue now is with the quality of the motion in scenes with heavy movement. I get zero dropped frames, and the AVI version looks great, but the mpg2 version does not. That's relly where my issue lies. I'm trying to avoid using 2-pass VBR plus the highest motion search, but using them does seem to give good results.

I should say that I"m splitting hairs here, almost ... the output is VERY good; it's just not quite as good as the original broadcast, where motion is concerned.

[Bizuser]

...
As for premier and not being able to do MPG2 that really suprises me and I doubt it, you probably only need a mpeg2 codec installed (possibly in the tools section, take a look).

The codec you point to is a 'decoder'. Words from the site itself "... This MPEG-2 Codec will allow you to open MPEG-2(DVD,SVCD,VOB) files in TMPGEnc and in MS Media Player, plus a host of other programs. ". I need something that plugs into premiere that will encode mpeg 2, which is, of course, a much bigger deal. I believe this feature was added to later versions of premiere (I'm using 6) but my version only allows me to 'file/export timeline/movie/as a Microsoft DV AVI (NTSC or PAL). Or, I can do 'File/Export Timeline/Save for Web', and then do things such as quicktime, RealG2, etc and what appear to be two types of MPEG1 - MPEG 2x CDROM NTSC (352x240). This is being provided by a plug in from 'Cleaner 5 Ez'.

The Canopus codec you pointed to is also a playback codec ... from the page itself, ..."This free software CODEC will allow machines without DVRexRTPro, DVStorm, DVRaptorRT, DVRaptor, EZDV, or RexFX cards to play Canopus DV CODEC AVI files. "

The amount of time it takes to create new file is mostly disc write time, time it then take the same file and and make a duplicate file you'll see what I mean.

Copying the file takes a fraction of the time it takes me to save the file in Premiere; here's what I did as a test; open premiere; new project; import clip; drag clip to timeline; mark an in point and an out point; 'extract' the selected section (cutting it out); then file/export timeline back to disk. For a 1.3 Gig file, with perhaps 0.3 gig cut from it, it took 9 minutes to export. A dialog appears, 'exporting' ...frame x of y Est. remain ...'. etc. Is this not how I should be using premiere to cut out bits? Copying the file, by comparison, took 2 minutes.

Thanks!

[thecoalman]

...
As for premier and not being able to do MPG2 that really suprises me and I doubt it, you probably only need a mpeg2 codec installed (possibly in the tools section, take a look).

The codec you point to is a 'decoder'. Words from the site itself "... This MPEG-2 Codec will allow you to open MPEG-2(DVD,SVCD,VOB) files in TMPGEnc and in MS Media Player, plus a host of other programs. ". I need something that plugs into premiere that will encode mpeg 2, which is, of course, a much bigger deal. I believe this feature was added to later versions of premiere (I'm using 6) but my version only allows me to 'file/export timeline/movie/as a Microsoft DV AVI (NTSC or PAL). Or, I can do 'File/Export Timeline/Save for Web', and then do things such as quicktime, RealG2, etc and what appear to be two types of MPEG1 - MPEG 2x CDROM NTSC (352x240). This is being provided by a plug in from 'Cleaner 5 Ez'.

!

That mpeg link was auto generated by the board.... anyhow I'm out of ideas.. Good luck.

[Bizuser]

That mpeg link was auto generated by the board.... anyhow I'm out of ideas.. Good luck.

Ooops! I didn't realise it was auto generated! Is there a way to distinguish between auto-generated links and user created ones? I"ve been using several phpBB forums recently, but haven't seen this particular behavior before. Cool feature, I guess ...

Thanks for your help!

[thecoalman]

That mpeg link was auto generated by the board.... anyhow I'm out of ideas.. Good luck.

Ooops! I didn't realise it was auto generated! Is there a way to distinguish between auto-generated links and user created ones? I"ve been using several phpBB forums recently, but haven't seen this particular behavior before. Cool feature, I guess ...

Thanks for your help!

Can't tell that I know of, cool feature but as this has pointed out can be a problem. Usually if it's just a word it's auto generated but the poster can do it that way too.

[Bondiablo]

you can eliminate step 2 completely and personally i prefer to use scenalyzer live or windv or dvio for capturing but if you encoded to 29.97 fps interlaced and the motion looks jerky or flickers the main thing that will fix your problem is encode using inverse telecine in TMPGEnc (flicker prioritized usually works best) and convert to a frame rate of "23.976 fps (internally 29.97 fps) & 3:2 pulldown when playback".

[Bizuser]

you can eliminate step 2 completely and personally i prefer to use scenalyzer live or windv or dvio for capturing

Presumably I can eliminate step 2 as long as I use one of these other tools? I'm forced to use it at the moment because Premiere only outputs Microsoft DV, and that is incompatible with TMPGEng (and VDUB). So I presume the tools listed generate a more compatible output to start with? I'll check them out.

If you encoded to 29.97 fps interlaced and the motion looks jerky or flickers the main thing that will fix your problem is encode using inverse telecine in TMPGEnc (flicker prioritized usually works best) and convert to a frame rate of "23.976 fps (internally 29.97 fps) & 3:2 pulldown when playback".

Your assumption here is that I'm capturing film-sourced material, is it not? If the original were film, then I would agree that inverse telecine would help, but the test show I was using was a Seinfeld show, which I would not expect to improve with inverse telecine - correct?

Thanks

[Bondiablo]

MS DV is totally compatible with TMPGEnc and you're captured file is compatible withVirtualDub as long as you capture or convert to type2 dv and install a DV codec. As for the other, try it and see for yourself.

[racerxnet]

Your bit rate is too low for the fast moving scenes which require the higher bit rate. Try a test with CBR at 10,000 burned to a disk with high motion scenes and see if this solves the problem. Use a bitrate calcutator to squeeze as much info onto the disk as possible. VBR below 4,000 is not to good when trying to fit to much info onto the disk. It's a hapy trade-off you have to decide on.

MAK

[Bizuser]

Your bit rate is too low for the fast moving scenes which require the higher bit rate. Try a test with CBR at 10,000 burned to a disk with high motion scenes and see if this solves the problem. Use a bitrate calcutator to squeeze as much info onto the disk as possible. VBR below 4,000 is not to good when trying to fit to much info onto the disk. It's a hapy trade-off you have to decide on.

MAK

That makes sense. But in TMPG, the max bitrate I can select is 8,000. This is not an area I've looked at before; is there something I can do to override this? I've been testing at both CBR 7000, and VBR 0/4000/8000. The VBR, with a max of 8,000, was definitely better. So if I can increase this to 10,000 that would be good. My current 'target' show is a 1 hour documentary, and I'll gladly fill up the whole disk with it.

Question related to the DV codecs... I'm currently capturing using premiere to a type 1 avi file (Microsoft DV codec). I then convert to type 2 (Canopus DV codec). I PRESUME this introduces some degradation, and that capturing using one of the products that capture direct to Type 2 will improve the overall quality (assuming I can't avoid going to type 2, which is my current assumption, given that I want to stick with TMPG and/or VDub)?

[proxyx99]

For a DV material I'd suggest you stick with 6000-6500 avg. VBR (8500-9000 max, 2500 min.) or 8-9k CBR, Anything below that will give you the impression that it is worse then the original DV. Always use highest bitrate your project permits (cosidering the final disc size).

[bugster]

You can install the Panasonic DV codec. This will allow you to use DVIO or WINDV to capture, and in type 2 DV as well. These DV files will then be accepted happily by TmpGenc and virtualdub.

Don't know if it will fix the jerky motion, but anything free has got to be worth trying hasn't it

[Bizuser]

Just wanted to say that I'm now getting some very good results. I did a comparison test, capturing the same show (broadcast later in the day) using my ATI AIW card using Picvideo MJpeg compression, set to 20 (max quality). Ended up with a 24 Gig file (and 240 dropped frames on my 'dedicated for video', 2.6 GHz system, with SATA drives); processed the whole thing through to a DVD, and compared. The Canopus version had 0 dropped frames, and ... only involved a file of about 12 gig size.

Thanks also for the tip on scenalyzer live; downloaded it and am testing now.

And as for bitrate ... I realised I was using PCM audio in TMPG, by mistake - so I was eating up bandwidth with uncompressed audio. And, just to clarify, the 'jerkiness' was really a very subtle thing, only noticeable in very specific circumstances.

So thanks everyone for working through this with me!

One remaining question I have ... the video 'color' from both captures is relatively 'pasty' (weak, low contrast) relative to a VCR recording of the same show (recorded still earlier - this was a 9/11 documentary that was shown over and over). Where is the best place (what tool, etc) to correct this? Can I do this in TMPG? Now that I've got all the other stuff out the way, I'm looking for perfection!

One other minor question, actually ... if anyone cares to answer ... why is the resolution sometimes expressed as 704 x 480, other times as 720 x 480 - where are those 16 bits going on the width?

[Bondiablo]

You're not using the MS codec to capture that's just what is used to read the file. The ADVC is doing the conversion and then just transferring the DV video file to your hard drive. If you didn't or can't save as type 2 you can use the DV Type 1 to DV Type 2 Converter. Using the canopus converter does the same thing but also makes it so the file can only be read with the canopus codec. If you want it to only be read with the canopus codec but don't need to switch between type 1 and type 2 you can just change the files 4cc code to CDVC with the AVI FourCC Code Changer. None of these result in any quality loss from the original file.

[jimmalenko]

Considering that editing and filter use (to clean up the video) seem to be a need, you might consider the following:

Transfer to DV using ScenalyzerLive and the canopus compatible codec. Create an AVISynth script and load this into VirtualDubMod to do your edits, and to insert your edits directly into the AVISynth script when done. Load th is AVISynth script into TMPGEnc and encode away. At any stage you can use AVISynth filters and view the results during editing or by previewing before encoding. FulciLives' guide should be able to help you with the formatting of your AVISynth script. I found I had to delete the DoubleWeave.SelectOdd() line to get my field order right. I too was experiencing what looked like 4 outlines around moving objects, and it turned out to be the field order from my ADVC-100. The other thing to remember is test your output on a standalone player as the computer can be a lot more forgiving.

[Bizuser]

You're not using the MS codec to capture that's just what is used to read the file. The ADVC is doing the conversion and then just transferring the DV video file to your hard drive. If you didn't or can't save as type 2 you can use the DV Type 1 to DV Type 2 Converter. Using the canopus converter does the same thing but also makes it so the file can only be read with the canopus codec. If you want it to only be read with the canopus codec but don't need to switch between type 1 and type 2 you can just change the files 4cc code to CDVC with the AVI FourCC Code Changer. None of these result in any quality loss from the original file.

Bondiablo - can you give me a pointer to a place where I can read up on DV stuff? I've been dealing with conversion of PAL DVDs (region 2) to NTSC DVDs (region 1), and all the issues of frame-rate conversion that that entails, but this is my first real foray into video capture.

I presume the video image is captured by the canopus device, and placed into the video stream sent down the firewire encoded in some compressed format - 'format x'. This format is set by the canopus device and is not changeable. My capture software (eg, premiere or scenalyzerLive) 'receives' this stream, and - perhaps - has the opportunity to change the format 'on the fly' to 'format y'. At this point, I have a 'format y' avi file. Is this correct so far? If so, what are x and y in terms of 'names'?

And how can I possibly simply 're-write' the code 'name' in the stream and get away with that - if the formats are different, how can I 'cheat' and set the fourcc code to 'y' when it is in fact 'x'? Or, is the fourcc code in the video stream telling the playback device what to use, and it's OK to tell it to use something different, as long as it can read the source ...? Thus, perhaps, the stream is 'x', but a 'y' codec can read 'x' without any problem ... ??? Am I even close here ...? A pointer to a good solid technical explanation would be welcome! In the absence of that, I'll just google around ... Thanks!
<update: found this via google ... http://plaza.ufl.edu/cnewsom/dv/digitalvideo.htm - looks interesting! >

[Bondiablo]

The reason you can get away with it without losing quality in this case is because you aren't really changing formats, it's all just Digital Video. DV is the compressed format the ADVC uses. When getting DV on your computer from a DV cam it isn't really a capture, it's a transfer of digital information from a digital tape to your hard drive. The ADVC is kind of a cross between that and a true capture in that it converts an analog signal to DV then sends that DV stream to your system. In both cases your capture software is just saving the DV stream to an AVI file. Some software like Mainconcept MPEG Encoder can convert the DV stream on the fly but for the best quality, especially if you want to do any editing, it's best to just save the DV AVI file and convert after. Changing the fourCC code of the AVI file to CDVC just means it will only be read with the Canopus DV Codec rather than the MS codec or whatever other DV codec you might install to open the file in VirtualDub (I use the Panasonic DV Codec). Changing from Type 1 to Type 2 is more like demuxing a video, you're just seperating the video and audio streams not changing formats.

https://www.videohelp.com/glossary?D#DV

https://www.videohelp.com/glossary?F#FourCC

[Bizuser]

The reason you can get away with it without losing quality in this case is because you aren't really changing formats, it's all just Digital Video. DV is the compressed format the ADVC uses. When getting DV on your computer from a DV cam it isn't really a capture, it's a transfer of digital information from a digital tape to your hard drive. The ADVC is kind of a cross between that and a true capture in that it converts an analog signal to DV then sends that DV stream to your system. In both cases your capture software is just saving the DV stream to an AVI file.

This makes perfect sense now; I read the links you provided, and researched this last night. Very helpful. But I'm still a little unclear on the role of the different codecs. DV 'is' the format; so whether I record using my canopus advc 100, or a DV camcorder, or any DV device, the compression is the same (even the compression level is fixed; it's whatever compression is required to ensure the DV-spec bitrate). DV also specifies the way the audio stream is compressed (PCM).

DV does not specify 'how' the video and audio are encapsulated; they are typically written to AVI files, and there are two methods - Type 1 and Type 2, and the two types differ in how the A and V streams are multiplexed.

All this is good (I hope!), but so far, no mention of codecs.

So .... my confusion is, what role, or effect, or purpose, does a 'canopus' Type 2 codec versus a microsoft Type 2 codec versus a panasonic Type 2 codec play? If the compression itself is predetermined by the source to be DV, then the 'decompression' should be fixed also? I can see the role of a Type 1 codec vs. a Type 2 Codec - this determines how the audio and video streams are delivered to the application. What changes in the output file based on the choice of codec, and why would I choose one vs. the other, since DV is a 'standard'?

Is it that each different codec (canopus DV, panasonic DV, etc) represents each vendor's different methodology of implementing the standard? I may choose canopus because they wrote better algorithms to decode the DV standard stream, and/or they encode my source material into the DV standard in a better way than the other vendor? So I can use Canopus DV, Microsoft DV, Panasonic DV; they all give me what I need, but they all do it in their own way, which may be faster or better?

OR, is it that each codec outputs differently formatted data; the input is standard (DV) but the output is whatever the codec writer wants to produce, and my choice of codec depends on which application I plan to use to process the data? I can choose Canopus Type 2 DV if my application supports it, but if it doesn't, I must choose another codec?

Thanks again!

[Bondiablo]

The codecs themselves aren't type 1 or type 2, any of them can read either type of file. The reason you need to install another codec to use VirtualDub is the MS DV codec is DirectShow only and does not suppport encoding or decoding for VFW and the reason you must save or convert to type 2 is, type 1 can't be used with VFW based editors. (you actually can open a type 1 file in VirtualDub but you'll get video with no audio) The role of the codec is to decode the file when playing/opening it and encode the new file when you save back to DV from your editing program. There are some differences in the codecs but it's just a matter of that vendors method of implementing the standard. I have seen some comparisons of the different DV codecs but this is the only one I could find with a quick search.

https://www.videohelp.com/glossary?V#VfW

http://forum.doom9.org/showthread.php?s=&threadid=58110

[NightWing]

If all would like I can put up on one of our web sites a split picture between Canopus decoder and Microsoft.

Actually the Microsoft one look richer, not as fuzzy. But both are about the same on high movement. About 80 to 100% movement. The difference is shocking on gray sky or shots with detail like a park with trees.

Did the "Split" bit to get handle on testing a ADS Pyro A/V link. The split idea get the playing field level so to speak.