How to Edit ATI Real-Time MPEG-2

[Nelson37]

This should also work for any VBR Mpeg-2 file, but is primarily intended for those using ATI cards and MMC, which produces files which are difficult to edit and maintain audio synch.

Capture file and rename to FILE.MPG.

Open M2-Edit Pro. Go to Panels, make sure Navigator, Composer, and Storyboard are open. Go to File, Open your captured MPG-2 file. Your file will appear in Storyboard in eleven segments. Now, the first segment always gives an error so you must start your capture approximately 10% early, timewise.

Start with the SECOND storyboard clip, drag it to Composer, go to File, Generate. Name it CLIP_1.mpg. Then clear Composer, repeat individually for all 10 clips in Storyboard. DO NOT EDIT IN M2.

Open TMPGenc, go to MPEG Tools, Merge and Cut, SVCD setting. Add File, hold CTRL and highlight all 10 files in reverse order, this will list them in the Cut window in correct order. Edit each file for start and end points, this may require loading some clips twice to complete the editing. Name the output file and hit Run. Voila, perfect synch.

Am still looking for a way to do the initial split with freeware, I'm not sure if TMPGenc gives correct synch because M2 has changed the file in some way or just because the file is smaller. Using TMPGenc alone does give correct synch but cut points can be off by 30 seconds or more, depending on file size.

[Nelson37]

Really surprised no replies.

Anyway, have also gotten success using ATI's Real-Time IVTC, which does seem to work. After the Storyboard step, you need to Demux the file, run Pulldown.exe, and remux. Then join as before. Still some issues to work out but this is giving VERY good quality @ 352x480, 2.6 bitrate selected, filesize is indicating 60-70 minutes per CD. Will update with final results.

The ATI IVTC puts the pulldown flag in the file, but M2 removes it, resulting in OOS. Replacing the flag after Storyboard restores audio synch. Results also seem to indicate that MMC is doing some kind of CQ encode as making no other change but adding IVTC results in significantly smaller filesize.

Would really like to hear from others using ATI real-time MPG2 capture.

[Bizuser]

I was going to do more testing before I posted, but as you are hungry for replies... ...

I've been testing mpeg 2 capture using ATI AIW cards. I first tried MPEG1 (for VCD), and had some success. Based on other posts in this forum, I was able to get a usable VCD from my MPEG1 capture. I did have to demux/remux in TMPG to correct some problems.

I then moved to MPEG2, for SVCD. The captured MPEG2 file itself plays great on the computer - no audio synch problems at all. But when I authored this mpg file to SVCD using VCDEasy, the audio synch was poor - as the show progressed, the audio got further and further behind - off by about 3 seconds after 30 minutes. Based on my success with MPEG1 using demux/remux, I tried this on my MPEG2 file; first, demux/remux in TMPG. No good (same problem). Then tried demux in TMPG, remux in bbMPEG. Lots of errors reported by bbMPEG (underflow...). Same output problem (audio sych drifts).

I'm interested in learning 'why' this is happening. I've been using the dvd2svcd suite to 'rip' DVDs, and in the process learned a lot about the process. I can now get 'perfect' rips but it took some time!

I'm guessing there is a tool that will 'fix' the MPEG2 stream produced by ATI, but have not found it yet.

Also - I'm using an old ATI AIW Rage 128 pro; I was looking at the new Radeon boards, and would happily spend the $$ to get these if they would do a better job, but I suspect the problem is in the software not hardware (and the quality of the capture on my 'old' board is pretty good - like I said, the mpeg2 file produced looks great on my monitor!

I'm willing to spend time and $$$ making progress on this - what should I try next?

Thanks!

[edsmith77]

bizuser,
I am also using the ATI 128 Rage Fury Pro card and it works just fine for me. However, I have not found an adequate (repeatable) method for splitting mpeg2 files yet. Mpeg1 files are quite easily done with TmpGenC.
After a bit of poking around inside the ATI generated mpeg2 file, there are several different audio streams created, the actual stream, a padded stream and sometimes one other. No amount or demux remuxing with different combinations would seem to make a difference.
I would be greatly interested if you found a repeatable method for splitting mpeg2 files where audio sync is maintained after the split or for anyone elses solution, if there is one.

Ed

[fingernailX]

I use an AIW 128 to capture @ 15Mbps I frame only, trim with Womble MPEG II VCR to individual clips and batch load into MainConcept Mpeg Encoder to quickly transcode to either DVD-R or SVCD format/size. I encode to one file I bring into Ulead DMF1 and create the video_ts folder which I burn with Prassi. Never a sync problem - never a hiccup - quality rocks and MainConcept is very fast! They are not freeware solutions but trials are available.

[Bizuser]

Ed:

I am also using the ATI 128 Rage Fury Pro card and it works just fine for me.

That's good to know! I'd rather not buy a new video card unless I have to, as it's always a chore to get everything setup cleanly.

Could you elaborate on what is working for you - what resolutions, formats, etc, and how you are processing the file, and what your output is (eg - 'capturing at xx res, mpeg2 format, authoring using VCDEasy - SVCD settings ...' etc )?

However, I have not found an adequate (repeatable) method for splitting mpeg2 files yet

I had the same problem - I created a 900 Meg MPEG2 file with ATI, and used TMPG's tools to split the file into two. No errors during the split, but ... the resulting files were hopeless. Even just playing them in ATI's file player, the audio/video starts out OK, but quickly gets out of synch, then ends up way out of synch, and finally the video becomes VERY jerky. Is this the same behavior you are seeing?

After a bit of poking around inside the ATI generated mpeg2 file, there are several different audio streams created, the actual stream, a padded stream and sometimes one other.

If that's true, then what about DVD2AVI? It lets you pick the audio stream to work with. I've been using DVD2AVI as part of the DVD2SVCD package, and it's very powerful. I'm quite comfortable using DVD2AVI and Avisynth as a frameserver combo to TMPG. Despite the name (...2avi), no avi file is actually produced so the file sizes don't grow. I'll give this a try this weekend.

I'm going to assume that, once you create a 'good' mpeg2 stream, TMPG's split tool will work OK with it.

fingernailX:

I use an AIW 128 to capture @ 15Mbps I frame only, trim with Womble MPEG II VCR to individual clips

Good to know! What made you choose I frame only? Is it an essential setting in your process, or just a 'nice to have'? Why 15 MBps - that's higher than even DVD rates. Is this simply so you have good material to start from?

I've had great success ripping DVDs and creating SVCDs, using DVD2AVI; Avisynth; TMPG for video encoding, beSweet for audio encoding, bbMPEG for muxing, etc. I'm assuming once I get a decent source file, properly encoded, I could use 'my' method for the latter stages rather than yours. So the challenge is, getting a good source file. What res are you using - actually, what settings are you using overall in the ATI capture stage?

I've read about Womble. It's $250, which is too much for me - I only want to copy a few VCR tapes - I'm not planning on doing this a great deal. I don't mind buying software, though, if there are other non-free solutions out there.

[furball6969]

TMPGEnc is quite good at splitting/demuxing Mpeg2 files.

It does tend to freeze up wehn working with large/long files, like 10 - 12 hour caps.

[Nelson37]

On the newer ATI cards. SFAIK the capture hardware is identical, no upgrade from the AIW 128 (except for games).

The extra padding stream ATI adds is the key, I believe it is related to TimeStamp problems which are inherent in Capturing. The key I have found so far is M2-Edit's Storyboard and Generate functions, you can even edit the source file as well as long as each clip is run seperately thru StoryBoard and Generate. Then join with TMPGenc.

Something M2 is doing, does not REMOVE the padding stream but seems to correct the video and audio so that other progs can work with it when they DO remove the padding stream.

After more testing with Settings, am now using 352x480, 2.5 MBPS VBR, IVTC, Motion 100, Standard GOP, Visual Masking, no Deinterlace. On Digital Cable movie caps, this gets almost 90 minutes on a CD and I can edit with TMPGenc, but I have to set start and end points about 30 seconds or more from where I want them or audio cuts out early at the end, the beginning is just later than set. Audio maintains synch and standalone plays OK, though.

DO NOT USE MMC PLAYBACK FOR SYNCH CHECK!!! WinDVD is far more accurate, or burn a cd-rw for standalone. MANY files which played OOS in MMC FilePlayer where just fine in WinDVD or Apex standalone.

[edsmith77]

bizuser,
Sorry for the delay in getting back to you.
I mainly capture at 352x480 in MPEG1. At around 2.5 Mb/s from VHS tapes. Depending upon the VHS length. I adjust the capture rate to match the size of the VHS tape so that I can get more onto fewer CDR discs. But it generally is in the 2.5 Mb/S range. I capture the whole file at this rate. This I load up into tmpgenc and split into however many discs is needed (I determined this before the capture was begun). THen turn tmpgenc loose. I personally use vcdeasy for authoring. Great program. It has worked flawlessly for me everytime. The only coaster I made was my own fault. Just careless fatfingering.
I capture to MPEG1 so that when I split the file, the audio and video remain sync'd. I used to capture to MPEG2 but never could get the audio and video in sync after the split. And I did not want to spend the reencode time for tmpgenc to make it all in sync. Otherwise I would have caputred in avi and went that route.
I have tried all of the default MMC 7.6 capture modes available. They all work great. Save one and that is the full DVD capture. I just drop to many frames. CPU speed or buss speed is not quite good enough to keep up. Since conversion from VHS tapes to full DVD is way overkill for me, It does not bother me not being able to go to that resolution.
However, I am still fiddeling around with changing the resolution and then doing some clipping to get rid of vcr border noise. Using the normal clipping within tmpgenc (or even mmc) does not work for me and my standalone. My standalone likes to have nice even numbers to work with. So capture at 352 then clip off 8 rows from the bottom to reduce the vcr noise results in a funny looking picture on the TV. Since the DVD player thinks it is still an even multiple, it ends up using left over or borrowed rows to keep the frame its same uniformity. Howeve, the picture on the bottom edge is quite screwey. But the rest is very nice. Have not found an elegant solution except maybe capture at say 360 then clip 8 rows to end up with 352. No results yet worth posting.

Ed.

[Bizuser]

Ed:

I mainly capture at 352x480 in MPEG1. At around 2.5 Mb/s from VHS tapes. Depending upon the VHS length. I adjust the capture rate to match the size of the VHS tape so that I can get more onto fewer CDR discs.

This resolution and bit-rate is not VCD standard. So are you just burning these discs for use on your computer? Or are you able to burn a VCD and play it standalone, even though your settings are non-standard? I was under the impression that VCD had to be 352x240, 1150 kbps.

I've read elsewhere that for VCR capture, you are wasting your time going above 352x240, 1150 anyway. Have not tested this myself yet.

I've been capturing live cable feeds at 640x480 (AVI, using MJPEG), then resizing them to svcd standard 480x480 in TMPG, with good success with VDub. I'm guessing that 640x480 is way more than I need for cable source, but then resizing 'up' introduces more artifacts so I don't want to capture lower.

[edsmith77]

Bizuser,
First, I am not an expert in the video reproduction environment. What I know I have learned through this forum and from the advice and experience of those here that are experts.
Now, you are correct in that 352 is a NON standard VCD resolution. In the parlance of this forum, it would be called XVCD. Not only for the resolution but the bit rate as well.
Yes, I am burning these for use in my standalone player. It has taken a bit of experimentation but my player can handle XVCD. However, it cannot handle SVCD. Go figure. But it can play DVDs quite nicely. To me SVCD and DVD mpeg2 should be the same the only question is the resolution and bitrate. Still do not understand this completely.
In any case, that is what drove me to mpeg1. I started out at 320x240 at 1150 bit rate. But was quite disappointed with the quality. So I upped the bit rate to 1500. Lo and behold my player was not disappointed with that bit rate. So I got bold and went 480x480 SVCD but alas it did not like that at all. Just would not read the disk or anything.
So I dropped back to xvcd 480x480 at 1500. Well it played that but the picture was awful. Not even viewable. But it did play. That is when I started playing with the resolution. 320x240 at 2.5 Mb/s looks pretty much the same to me as 320x480. Thus I tried upping the horizontal number to a point where the picture became garbled. Anyway the max for me came out to be 352x480. I have even ran this at bit rates up to 8Mb/s with out any problems. Except at this rate, you can only get about 15-20 minutes on a disc. To me that is just not worth it. So my compromise was to stay around 2.5 Mb/s which gives me about 40 minutes of video and an audio rate of 224. This makes me happy.
It aint DVD ! but it is more than acceptable to the casual observer.
As far as a waste of time. I think that is in the eye of the beholder.
As far as resizing and artifacts, the experts here, for the most part, will tell you to capture at the final resolution rather than resize later. Not sure why other than unnecessary processing can introduce "artifacts" or other unwanted things. I always try to capture at my final resolution. Just makes logical sense to me the non expert.
Hope, this helps. If I can pass along any other help, I will be glad to offer.

Ed

[oldfart13]

I cap in high Mpeg-2 (6 MB/s +) 480*480 CBR, check sync in dvd2avi and if it's good, load file into MyFlix XE, edit out commercials and save result. Convert to VCD or SVCD with TMPGEnc, Procoder, etc. Results of late have been great and in sync. If the sync is out after capping, I correct using the method on this site with Goldwave. Takes only a couple minutes to demux, correct/normalize, mux. Been using Multimedia Center 7.6 for the past several months. Works great for what I am doing...

[Bizuser]

Thanks for the info Ed.

320x240 at 2.5 Mb/s looks pretty much the same to me as 320x480. Thus I tried upping the horizontal number to a point where the picture became garbled. Anyway the max for me came out to be 352x480.

If 320x240 looked the same as 320x480, did you try 352x240 instead of 352x480? I suspect you'd get the same results VCD spec is 352x240. Once you go above 240, I think the playback in a VCD player doesn't know what to do with the extra info - so you may be wasting 50% of your bandwidth.

I never thought about XVCD - I could try that. What player do you have?

The only other drawback I can think of is 'future compatibility'. If your player goes up the spout, a replacement player may not play your non-standard VCDs. My current thinking is that the whole VCD, SVCD thing is pretty flakey. I just burned a great, standard, VCD that played on 3 players at home (all sub-$100), I loaned it to a friend who had an expensive Sony player. The audio was off by 2 seconds or more right off the bat! So I ended up recording it to VCR and loaning her that!

[Nelson37]

352x480, w/mpeg-2, is a standard - CVD. It is also valid for DVD.

The reason for capping VHS at 352x480 is the interlacing, if you cap at X240 you are not using half of the vertical data. For target of VCD, resizing later uses both fields for a result which theoretically should be better, users choice, of course.

Player compatibility is always an issue, I lucked out that the el cheapo player I got was an Apex 500 which plays almost anything. Players which do not play SVCD are perfectly capable of doing so, it just recognize the disk. There is a physical limitation of data density (bitrate) due to low spin speed - DVD's are more densely packed than CD - but many players can handle bitrates over 4000.

EdSmith, how are you clipping the video ? Or are you re-encoding after the clip?

[miamicanes]

On the newer ATI cards. SFAIK the capture hardware is identical, no upgrade from the AIW 128 (except for games).

Almost I think the AIW 128 Pro (specifically, the retail, or maybe just the 32-meg version) was the first AIW to have the Rage Theatre chipset.

I'm 99% sure the AIW 128 (non-pro) didn't have it, and I'm about 85% sure that at least one of the non-retail AIW 128 pro configurations didn't, either. If my memory is right, there were three AIW 128 pros, available in both AGP and PCI

* AIW 128 Pro, retail (32 megs SGRAM, available as AGP or PCI)
* AIW 128 Pro, OEM (32 megs SDRAM, AGP or PCI)
* AIW 128 Pro, OEM (16 megs SDRAM, NO Rage Theatre chip, AGP or PCI)

At least, that's what I remember from the research I did back in Spring2k before buying mine. I just remember that there were two models that were close (differing w/regard to ram), and one model that surprised me ATI could even legally get away with CALLING it an AIW 128 pro

<flame>
Another thing that pisses me off... the practice of selling blatantly inferior cards under the same name, but "OEM". Even if they aren't supposed to wind up in retail channels, OEMs like DELL and Gateway shouldn't be allowed to claim their PCs have (insert name of high-end card here) without having to prominently disclose EXACTLY how their version deviates from the retail version reviewed in magazines. After all, it's the reviews of THOSE cards that buyers are basing their decision on.

Anyone care to speculate how many seconds it would take a Grand Jury to indict a car dealer if it tried selling Escort GTs (sight unseen) as OEM Mustangs (with Ford's knowledge and encouragement)? Taking orders via mail order, with a brochure full of Mustang specs and photos, and nary a word anywhere (including the order form) as to how, exactly, the OEM version differs from the one advertised on TV and reviewed by auto magazines?
</flame>

[Bizuser]

Miamicanes - I've always been confused by ATI's products; your post made me realize I'm not the only one! I have the retail box "All In Wonder 128" 16 MB PCI Bus. Also has a logo 'graphics by rage 128'.

When I go to the Device Manager in win2k, it shows as 'All-In-Wonder 128 PCI'. When I go to 'details', I see 'ASIC Type: Rage 128 PRO'; Tuner Type Philips ...; Decoder Type: Rage Theater ...

So is it 128 or 128 Pro??!!! I guess I don't care, but it makes choosing drivers pretty hard!

[drewson99]

This should also work for any VBR Mpeg-2 file, but is primarily intended for those using ATI cards and MMC, which produces files which are difficult to edit and maintain audio synch.

Capture file and rename to FILE.MPG.

Open M2-Edit Pro. ...

I kinda stopped reading here. M2 costs a fortune. i have given up on capping in mpeg2, now I use a D8 VCR>IEEE1394>AVI>DVD.

i recommend this method to all who want their capped tv/cable to look like the original signal when watching it on DVD.

ZZ

[miamicanes]

So is it 128 or 128 Pro??!!!

My guess is that it's an AIW 128, but then ATI rolled out the 128 pro, had a ton of 128's left, and decided to let them be sold as OEM 128 pros -- exactly the kind of thing that, if true, absolutely astounds me that they've never been sued over.

Think about it:

* AIW 128 pro was explicitly advertised as having hardware MPEG-2 capture. And, in fact, we all know it *does* have hardware to accelerate the process.

* HOWEVER, no driver that ATI has *EVER* officially released for the AIW 128 pro, for *ANY* operating system (*INCLUDING* Win98 SE), has *EVER* actually taken advantage of it.

I think it's safe to say that if push came to shove in a courtroom, ATI would be told to either

* refund the purchase price of every AIW 128 pro sold,

* offer free exchanges for the lowest available Radeon that *does* perform as the 128 Pro was advertised, or

* make official drivers supporting hardware-accelerated MPEG-2 capture *officially* available *NOW* -and- pay damages to all the AIW 128 pro owners who paid $200-350 for a video card that has officially never worked as advertised in the three years since it was sold.

I agree that hardware marches forward, but the way ATI specifically handled the whole AIW 128 Pro issue -- dropping it like a hot potato the nanosecond the AIW Radeon hit the market, even though the 128 Pro had *identical* capture hardware and could have been *trivially* kept up to the same level of support with respect to it, was just plain wrong. The 128 pro wasn't some $16 bargain bin card from some nameless manufacturer in SE Asia. It was their flagship product before they decided to take on the 3D gaming universe and focus all their efforts on boosting their QuakeMark score instead of making the video capture work right.

Sadly, ATI is by no means alone. Among graphic card manufacturers (with the possible exception of Matrox), it's more or less the way they do business. But the 128 Pro case is particularly galling, because it represents an example of a card whose subsystem of primary interest to the card's purchasers (capture) was, in fact, architecturally identical to the cards that replaced it and could have been supported with *minimal* effort on their part. Hell, the fact that a motivated user *can* manually rip apart the driver bundle to an AIW Radeon and perform his own guerrilla update is proof of it.

[edsmith77]

Nelson37,
I have given up for the time being on trying to solve the clipping problem. Therefore, I just ignore the vcr noise bands at the bottom of the screen. As I said earlier, I did not want to invest the 8+ hours of reencode time to try to get rid of those pesky lines. Hence, I just mentally tune them out. Most of the time they only occupy the lower 8 lines or so of hte screen and are not that noticeable. However, on some clips they are really bad. Sometimes they move up into the 30-40 row range and cannot as easily be ignored, in that case I just dump the original as a lost cause and move on.
Now that I think back a ways, there used to be (on the older vcrs, anyway) a mechanism to adjust the tracking which would almost always track the noise off the screen so that you just could not see it. However, the newer vcrs (or mine anyway) do not have anyway to adjust the tracking. So my choices are to ignore, dump or pay the reencode price.

Ed

[theboss777]

What verison of m2editpro are you using? I am using version 5 and when I open up the mpg, it does not split into segments.