Frustration setting chapter start points

[C53248]

I am used to making (X)(S)VCDs. With the VCD's it was enough to drag the mpg into Nero and burn it. Authoring is a new pain in the butt.

I have a 4GB DVD-compliant MPG file of about 2 hours 20 minutes. I want to have chapters start at three places (including the beginning). I am finding it terribly frustrating to set the chapter starts that are roughly 1/3 and 2/3 into the video. DVD Workshop and DVD Movie Factory have a straightforward interface to do it, except that moving through the video is really slow, and getting it to a particular frame is painfully tedious. When I move the slider, it will only move so far, and then I have to wait for a while, and it eventually shows that frame - just extremely unresponsive. I'm used to the sliders in VirtualDub and TMPGEnc that respond in real time. Is it this difficult, or am I doing something wrong, or do I maybe have some performance bottleneck (256 MB RAM, 1200 MHz Athlon, plently of GB's HD, Win98SE)?

[igalan]

I have the same problem here . DVD WS is too slow to set the chapters in a M2V file larger than 1 GB. If someone knows a trick to do so... But I guess that we'll have to wait for an update. I have AMD XP 2000+, 384 MB DDR RAM 266 MHz, Windows XP Home, 40 GB 7200 rpm disk.

[digitalmaster]

When using DVD Movie Factory, instead of using the sliders, click on the seconds,minutes, or milliseconds in the time box. Then click on the up or down arrows next to it. This will let you move forward or backward by milliseconds, seconds, minutes, or hours. You cannot get more precise than that.

[Coop]

C53248 & Igalan,

I've had the same playback problem with setting chapter points, but the hesitation/offset you refer to doesn't happen for me until halfway through the clip. I've found that the actual counter time & frame display is off - when I burn and playback it seems to be anywhere between 6-20 seconds off from where it set the chapter point versus where I intended it to be.

How did you encode your MPEG? I used TMPGenc 2 pass VBR with Optomized settings, and I'm thinking this has something to do with it. The first project I did was using CBR and I didn't notice any issues with playback or chapter points.

[igalan]

How did you encode your MPEG? I used TMPGenc 2 pass VBR with Optomized settings, and I'm thinking this has something to do with it. The first project I did was using CBR and I didn't notice any issues with playback or chapter points.

I tried this with a ripped DVD, I used the ripped M2V (~3 GB) in DVD WS. I don't have any problem with the MPEG2 movies that I capture, even if they're over 1 GB.

[Coop]

Igalan,

I'm wondering if it's not so much the file size, but rather the encoding specifications. Are you able to control the MPEG2 encoding specs when you rip?

[igalan]

Maybe it's the encoding specifications, this is something to try. With all DVDs that I tried, I had this problem. Smartripper all it does is remove CSS and Macrovision, it won't change the encoding at all. I will try with other movies to see what happens. I am experimenting if it's possible to back up DVDs with DVD Workshop (which you can), but adding chapters has been impossible.

[davideck]

I am having the same experience and thought it was my PIII 866Mhz system. Anyway, I get around the problem by first burning a DVD-RW in fast editable mode with chapter points every 3 minutes. Then I watch the DVD-RW in my home player and write down the timecode of the places where I want chapter points. I can also verify that audio and video remain in sync. Then I go back to MF2 and type in the timecode of the chapter points by hand. It may sound tedious, but I find it much more enjoyable than trying to manipulate the slider bar.

[Coop]

david - that's how I got around it as well, it took 3 burns or so to get the chapters modified correctly (counting the offset of seconds and then adjusting the chapter point setting) as the only way I could find out if I did it right was to watch the DVD and scan the chapter points. I'd like to know why this is happening though, as that adds 3 extra hours of time to keep recreating, burn, and erase.

[miamicanes]

Same problem here... here's what I've observed:

* it only happens with VBR-encoded files.

* the delay increases in a linear fashion along with the distance from the start of the clip. Near the beginning, there's basically no delay. Every 60-90 minutes seems to add 25-40 seconds on my Athlon 2100XP w/15k Cheetah & 512 megs.

Putting 2 and 2 together, I think it's obvious that the program is literally parsing the entire file from the start every single time it needs to find an I-frame.

It's a shame, because it basically renders VBR and DMS2 mutually unusable with 60-120 minute video clips.

What's tragic is that it would realistically take one of their programmers maybe a couple of days at most to implement a parallel I-frame index for VBR MPEGs... storing the timecode and byte offset of each I-frame so that all it would need to do is look in the index to find the nearest timecode, obtain the corresponding byte offset, and go directly there to retrieve the I-frame from the MPEG file to show the preview.

[C53248]

I see I'm not alone with this. Thanks for the suggestions, insight, and such.

Yes, in my case it's a VBR MPEG. I captured it from VHS and encoded it with TMPGEnc using 2-pass VBR (2000, 3700, 8000). My other encoding settings are mostly ordinary for DVD. It makes sense to me that it's the VBR. And I'd like those programmers to get cracking on that indexing too. But I thought the MPEG files are supposed to be smart enough that the player can just jump anywhere. The stand-alone players don't have this problem.

The precision of the time display is good enough (accuracy perhaps another matter) along with the ability to move to the next or previous I-frame. But it's tedious if you are looking around for the place to put your chapter. In my case, I'm looking for where one episode ends and the next one starts, and there's a particular I-frame where that happens.

I tried to use PowerDVD to find the spots, but it doesn't agree with the authoring programs about the time. I think it was off by minutes. I wonder which one is right or if either one is. PowerDVD also gets bogged down scanning for a certain spot in the video, though it behaves a little differently. In PowerDVD I had trouble jumping past about the halfway point of the video. From there, the best I could do is play forward at 8x speed.

I may try encoding in CBR instead. I don't like thinking about what that does to quality, but this is too much work.

[miamicanes]

I think it's because standalone players don't need to care exactly WHERE they are, as long as they can find something in the vicinity to play. In other words, when somebody holds down the search button to scan forward, the player can get away with jumping ahead by some interval and looking for the next I-frame. It doesn't have to say, "Hmmm. OK, right now I'm at 1:07:03.15, and I need to skip ahead by exactly 109 fields to get the right cadence..." It just needs to say, "Hmmm. If I just jump one groovewidth towards the center and show the next I-frame I find, the user will be happy.

[G-man]

Try this freeware utility: http://stevemiller.net/downloads/MFScene.zip

It runs alongside Moviefactory and makes it much easier to move thru the video looking for chapter points. Works with both MF1 & MF2.

[dvcd]

Dvd-workshop 1.0 can set chapter start points very fast.
Do not use 1.2 or 1.3.

[kitty]

You might select Close GOPs in your encoder (like TMPGenc) - this will probably allow better scrolling through the file at the expense of a little increase in size.

[C53248]

You might select Close GOPs in your encoder

I did encode with Closed GOP, so I can say that didn't help.

[C53248]

G-man,

I tried that MFScene. Everything in its window is ghosted, and it's not interacting with MF2. I just put its files in MF2's program directory and ran it. Is there something else I have to do to get it going?

[G-man]

I think I mis-understood your question. MFScene works for selecting thumbnails to use for the chapter points, not actually selecting the chapter points themeselves. Sorry about that.

If you start up MFScene and then click on a thumbnail (when creating your menus) in MF2 you can then control frame location with MFScene.

[SaSi]

Your problem appears to be caused by GOP structure. Closing COPs will not help unless you have them. I was making the same mistake after doing a lot of SVCDs and then starting DVDs.

If you are using tmpgenc, try the following settings for GOP structure:
Number of I pictures in GOP = 1
Number of P pictures in GOP = 5
Number of B pictures in GOP = 2
Output Interval of Sequence Header = 1
Max number of frames in GOP = 15

I think you are not using the last two parameters correctly. If the last parameter is set to 0 in Tmpgenc, then your movie may not contain many I frames. If the previous value is 0 (the default) then your whole movie is going to be a huge GOP and random access is virtually impossible.
In case of doubt, try to encode a file with your settings and press CTRL-L for the conversion log to appear on screen. You will see the exact structure with I, P and B frames.

The smaller the GOP length, the better random access you will have.

PS. I had a similar problem when previewing MPEG-2 files with Media Player. Moving the position slider to the midle of the movie caused the PC to freeze and not respond (it was scanning the stream doing decoding as fast as it could) for minutes. Using the settings above solved all my problems.

PS.2 If you want to have a nice and almost complete introduction to MPEG, visit
http://bmrc.berkeley.edu/frame/research/mpeg/mpeg2faq.html

[C53248]

G-man,

Thanks. I'll try MFScene next time I'm selecting the thumbnails.

Sasi,

I'm using those default numbers: 1,5,2,1 and 18 is the maximum GOP length, though with a closed GOP, it only gets to 16 before it starts over. The encode log does show that pattern, IBBPBBPBBPBBPBBP, with 16 frames per GOP throughout the video.

dvcd,

Dvd-workshop 1.0 can set chapter start points very fast.
Do not use 1.2 or 1.3.

Thanks. I tried that, and you're right. DVDWS 1.0 is not nearly as bad. It's still a slow process, but now I can imagine doing it more than once. I can keep doing VBR.