Different Clips w/Different Bitrates for DVD?

[Quamatek]

I have several clips of video I'd like to put on a DVD. I already know that some clips will be encoded differently based on how the sources were encoded (e.g., 29.97 fps Progressive, 23.976 Progressive, 59.94 Progressive). My question is about the bitrate settings for each of the clips.

Some clips have noticable better quality than others and I'd like to maintain as much quality when do the encode for DVD. Should I:

1) Total the clip times together and calculate the bitrates for all (so the bitrate settings are the same for each of the clips)?

or...

2) Encode each clip with different bitrate settings based on the source quality?


Additional information:

Some clips will contain PCM audio while others have Dolby Digital 2.0 (so that may impact the overall bitrate). I've allocated about 200 mb for menu just in case. I'll use DVD9 for larger-sized clip projects while DVD5 for smaller-sized clip projects.

If you need more information please let me know. Thanks!

[guns1inger]

Using PCM audio will indeed impact the video bitrate - substantially. Unless there is an absolute need for it, 384 kbps AC3 will probably be more than adequate, and will give you more than an extra 1 Mbps of bitrate to play with.

I would divide the clips into order of importance, then within that, quality. Work out how much space you are prepared to give up to the important material, then start working out bitrates.

[manono]

1) No.
2) Yes.

Best of all, perhaps, would be to encode each using a constant quant. That way they'll all get the same quality, or have the original quality diminished equally. You don't want to give each clip the same bitrate as different clips have different original qualities and will compress differently.

When I have different clips to put on the same DVD I'll first encode them using CQ, and then, if necessary, do multiple passes afterwards, assigning different bitrates to each based on the percent of the total file size of the original CQ encodes. Either that or adjust the CQ and/or quantisation matrix used. I hope I'm making sense. I'm using CCE, and this sort of thing is fairly easily accomplished when using that encoder.

Some clips will contain PCM audio while others have Dolby Digital 2.0

If it were I, I'd reencode any PCM WAV audio to DD 2.0 AC3. But if you keep the WAV audio, you might adjust the max video bitrate, if you like your max video bitrates high (as I do).

I've allocated about 200 mb for menu just in case.

That's a mighty big menu, but I guess you know that already. Me, if space is at a premium I prefer to devote as much space as possible to the videos.

[Quamatek]

Ah, thanks guys for helping.

I'll go with #2 as both of you recommended that I do. I initially thought that was the better option but after reading threads on bitrates I got confused whether or not the clips needed to have the same bitrate settings. By encoding the bitrates based on the clips' source I can ensure the best quality for each. I've prioritized my videos in the order of importance and at the same time will shoot to keep the better source as much as possible.

When we speak of "constant quant" are we referring to CBR vs VBR or Quantization? If Quantization, is there a recommended setting for DVD NTSC or will it depend on the source? If CBR, is there an advantage to first encode using CBR then a second encode of VBR with a 2-pass?

For audio I've set my A/V receiver to play PCM in 2.0 and Dolby Digital through my center channel speaker. I noticed that most of my music video DVDs have the audio encoded in PCM. The music video DVDs that were encoded in 5.1 used Dolby Digital. I think 384 kbs is more than enough and will give me more room to play with the video bitrate. When I compared the video bitrates I increased it from 8000 to 9200 with only a 5 mb file size difference. Does a 1200 separation make a difference in video quality or does it depend on the source?

Update: I'm experimenting with 9100 and 384 audio instead of 9200.

I'm not exactly sure why I allocated 200 mb for the menu since at this time it's taking up roughly 1/4 that amount. :P At any rate, I'm not squeezing for space since I only have 8 videos on one project with 1 gig to spare.

[manono]

When we speak of "constant quant" are we referring to CBR vs VBR or Quantization?

CQ or constant quant or sometimes known as constant quality is very definitely variable bitrate. It doesn't get any more variable than that. Recommended setting? Depends on the numbers used in whatever encoder you're using. You haven't said, have you? I often use 35 in CCE.

If CBR, is there an advantage to first encode using CBR then a second encode of VBR with a 2-pass?

Don't even think about using CBR encoding unless the total length of these things is only an hour or so. Even then I wouldn't use CBR. But plenty of people would.

[Quamatek]

Thanks for the clarification.

There's a VBR CQ setting for TMPGEnc Xpress. I have several DVD projects I'd like to do. Some will have videos that are 60 minutes combined while others will hit around 150 minutes. Let's start with one of the 150 projects:

I have two video clips for the 150 minute project. One video has better quality than the other. The source specs are:

Clip #1:
1:10:00 (127487 frames)
624x352 29.97 fps
Dolby Digital, 48,000 Hz, 2ch

Clip #2:
1:18:00 (140772 frames)
1280x720 29.97 fps
48,000 Hz, 2ch

----------------------

In a bitrate calculator (in signature) the settings are:




----------------------


Project #2 (60 minute DVD):


I'm fairly new at video stuff so this CBR vs VBR is confusing when the recommended settings are the same for the maximum, average and low. Case in point, for the DVD project that's roughly 60 minutes total (8 video clips; 1 is about 14 minutes while the 7 is between 5 - 6 minutes each). Here are the calculations below:




I'd like to use VBR if the video quality is better. I don't know how to set the average and minimum bitrate based on these results. If you need specific information please let me know.

[manono]

If you're asking me, I don't use TMPGEnc. Sorry.

[Quamatek]

Oh. I'm using this bitrate calculator to output the numbers. I assumed that CCE and TMPGEnc Xpress would both honor the calculated results the same way. :/

I'm confused how to reach the average and minimum bitrates using VBR CQ when the max, average and min are 9250 (second screenshot)? Based on the calculator results, it says to use CBR. I'd like to try your recommendations so any help is appreciated.

[jagabo]

Constant quality encoding is the flip side of VBR encoding. In VBR encoding you specify an average bitrate (and hence the file size, file size = average bitrate * running time) and the encoder delivers whatever quality it can for that bitrate. In CQ encoding you specify the quality and the encoder uses whatever bitrate is necessary to deliver that quality.

The problem with constant quality encoding is you don't know what the final average bitrate will be. See the videos in this post as examples (they are encoded with Xvid but the idea is the same):

https://forum.videohelp.com/topic357646-60.html#1903520

Both videos (the same source but one had noise added) were encoded with the same quantizer. The noisy one required 50 times the bitrate of the clean one.

The best way to get a specific file size and similar quality using VBR is to encode all your videos as one big video. If that's not practical you can encode each video using CQ encoding then look at the final file sizes. That will give you an idea what relative bitrate to use for each video.

[Quamatek]

Thanks jagabo. I saw your samples and noticed the clean version took up the least amount of space. So a VBR CQ will use only the necessary bitrate for a scene? If so, I can see why CBR is inefficient when you're trying to squeeze for room since it may use more bitrates than needed. If I understand correctly, even if I have disc space I should opt to use VBR CQ rather than CBR.

Unfortunately I'm not sure I can encode the videos into one big file because the encoding methods will be different. Some will use a 3:2 pulldown while others Progressive or Interlaced.

I see the quality setting for VBR CQ. I see what manono meant by "35". I ran 4 tests (only encoded video m2v) and in each I used the settings:

1)
Maximum bitrate = 9200
Minimum bitrate = 2
Quality = 10
Fize size = 15.9 mb

2)
Maximum bitrate = 9200
Minimum bitrate = 2
Quality = 50
File size = 27.5 mb

3)
Maximum bitrate = 9200
Minimum bitrate = 2
Quality = 100
File size = 78.1 mb

4)
Constant bitrate = 9200
File size = 84.7 mb


#1 there was heavy pixelation. The outcome was drastically worse than the source. With #2, the output was better with some pixelation. In #3 the output was close to the source. I didn't see much difference between #3 and #4 (cbr) during normal playback. Only the size differed by 6 mb. Considering this was about a minute video the file size will make a difference in the final DVD thus making the VBR CQ more efficient. Did I get it right? :P

I've been searching around and just when I thought I was beginning to understand CQ I see automatic VBR CQ vs CQ. Is the former VBR 2-pass while CQ is VBR CQ (where I'm able to adjust the quality)?

[jagabo]

Your first idea about single pass constant quality encoding is correct. The encoder uses only the bitrate necessary for each frame to achieve the requested quality. This mode often has a cap to prevent bitrate spikes from exceeding the maximum allowable bitrate for DVD (or whatever purpose you have in mind).

Single pass constant bitrate mode uses a fixed bitrate for every frame. Scenes which don't require the requested bitrate still get that bitrate so they waste space. Scenes that require more bitrate don't get it and the picture quality suffers.

Two (or more) pass variable bitrate examines the file during the first pass to see which parts need more bitrate and which parts need less. During the second pass it uses that information to apportion bitrate to different scenes and still meet the requested average bitrate (and hence file size). You use this method most often when you need to maximize quality for a particular file size. Like 700 MB on a CD or 4300 MB on a DVD.

There is another type of single pass VBR encoding where you select a bitrate and the encoder tries to meet that bitrate while still using more bitrate on shots that need it and less on shots that don't. But since the encoder can't know what's coming later in the video it can't make the best decisions as it encodes. For example, you may have a movie that consists of all clean still shots for the first half. The encoder will see that and start using higher bitrates than really necessary thinking the whole movie will be the same. But then the second half of the movie comes along and it has lots of action, noise, etc. Now that the encoder has wasted lots of bitrate on the first half it can't use as much as is needed on the second half. It either has to encode the second half with lower quality than necessary to meet the average bitrate requirement, or let the average bitrate rise to account for the high action video. The exact strategy varies from encoder to encoder.

Note that with most sources there's not much improvement going over 8000 kbps or so. Use a Bitrate Viewer to examine the bitrate of your files. You'll find it interesting.

[Quamatek]

Updated (after posting): I just noticed your screenshot. I'll check the bitrate viewer.


Great information, thank you. Encoding does get complicated.

Your first idea about single pass constant quality encoding is correct. The encoder uses only the bitrate necessary for each frame to achieve the requested quality. This mode often has a cap to prevent bitrate spikes from exceeding the maximum allowable bitrate for DVD (or whatever purpose you have in mind).

I've been testing 6 videos so far with the settings of:

Maximum bitrate = 9200
Minimum bitrate = 2
Quality = 100%

Audio = 224
GOP and Quantization I left at default.


Of the 6 videos the highest average bitrate is 9070 which is pretty good. I don't think there was a spike above DVD allowable specification. The lowest average bitrate was around 6500. The videos looked good during playback but I don't know if I got the most out of the settings. For these tests I'm encoding them individually. If I'm doing something wrong please let me know.

I'm still a bit lost about the minimum bitrate. I've set it at 2. I've read several times that some hardware DVD players have difficulty playing when bitrates go too low. Some recommend staying above 2000 - 3000 and I've read a couple of times where a setting of 0 is acceptable. I haven't tested it on my DVD players so I cannot say one way or another.

For example, if my max bitrate is 9200 with audio at 224 while the average is around 8600 should I set the minimum to 8000 (600 above and below the average)? Or should I leave it to 0? I've read where people say if the differences between the minimum, average and maximum are close together there's no benefit in VBR. What say you?

[Quamatek]

Wow! The Bitrate Viewer is awesome! I can pinpoint the locations of the bitrates and where it dips and peaks. Based on the encode it told me my minimum, average and maximum. One of the videos spiked to 10082. Something tells me I need to lower the 9200 max setting.

[jagabo]

Don't take Bitrate Viewer's bitrate peaks too seriously. The bitrate limit for DVD isn't for a single frame but rather the average over some short period. Bitrate Viewer lets you see the average over 1 frame, 1 second, or each GOP (usually about half a second). I don't know if either of the latter periods corresponds to the the DVD limit. On the other hand, I know a lot of people have recommended setting the max around 9000 or so.

I haven't used TMPGEnc Express. But I have used the older TMPGEnc Plus and the encoder appears to be pretty much the same. A quality level of 100 is pretty high. I hardly ever used more than 95, and more often around 90. You'll find bitrates drop quite a bit at those settings, without much change in visible quality.

Regarding minimum bitrates: I've also heard people say you shouldn't go lower than 2000 or so. I've never had a problem setting the minimum at 0. Of course, I've only tested on a dozen or so players. I have seen problems with very low minimum bitrates on SVCDs. Playback gets jerky on some players at very low bitrate shots.

[Quamatek]

So far even with the occasional spikes to 10,000+ the video still plays fine. I fed the m2v files into muxman and the program accepted them so I guess they are DVD compliant. I hope so. :P

I've been experimenting with the files and will post some specs after I'm done. Because the sources are not at the highest quality I wanted to keep as much quality as I can through the encode. So far when I left the quality to 100% and video bitrate to 9200 that rendered the best results. When I encoded with bitrates of 8000 I immediately noticed diminished quality. I could get by at 9100 and it was pushing at 9000. I've also tried 9200 with 95% and 90% settings and to my surprise, while there was a difference in quality, it wasn't as much as I thought would be. I left the minimum at 2 and so far no issues.

If the source of the video is low will I gain anything by encoding at a higher bitrate or will it just be wasted space? There's a saying like "Crap in, crap out" and it holds true with audio. While I assume it is, does it also hold true for video?

[jagabo]

when I left the quality to 100% and video bitrate to 9200 that rendered the best results. When I encoded with bitrates of 8000 I immediately noticed diminished quality.

I would check that again. Every source is different, but in my experience there is almost no visible difference between those two settings. Even when looking at enlarged still frames.

I've also tried 9200 with 95% and 90% settings and to my surprise, while there was a difference in quality, it wasn't as much as I thought would be.

With the quality setting at 95 to 100 percent you are telling the encoder to work really hard at preserving essentially invisible artifacts in the source. Again, in my experience there is little visible difference between the 90 and 100 percent settings. But the video encoded at 90 will be significantly smaller.

What kind of differences were you seeing? Be careful about comparing with media players. It's likely one will be using your graphics card's video overlay feature and the other not. Video overlay has its own proc amp settings and may include other features like noise reduction, sharpening, etc. This can result in the exact same video looking very different in two players playing side by side.

Also be sure you both frames are the same type. B frames are encoded with lower quality than I and P frames. The encoder may not always encode the same frame as the same type. If you are looking at an enlarged still frames, and that particular frame is a P frame in one source and a B frame in the other, you'll find the P frame looks closer to the original.

Here are a few samples I encoded at CQ 90 and CQ 100 with bitrate max 9200, bitrate min 0:
http://www.mediafire.com/?dscbxjs3bl3

[Quamatek]

Sorry for my delayed response.

You're probably right about having little visible difference at 90 - 95% quality. I've been using 100% thinking it may give me the best replica off of the source. While that may not hold true I'll have to experiment more.

Unless my eyes are playing tricks on me I can see some pixelation and color differences when I drop to lower quality percentages. I've been consistent testing it on WMP (on MPEG-2) then PowerDVD (after authored to DVD) then my main DVD player for the HDTV. I believe the source has a lot to do with it. What is the saying, "Crap in, crap out"? My goal is to preserve the video and audio as best as I can without filters whenever possible. I'm quite happy how they're turning out even with my lack in encoding experience. :P

I'm still confused about the I, P and B frames. The only thing I'm aware of is to stick with 18 and under. I've been leaving the default settings recently. Although I've read having the right values can make a difference I don't want to mess the videos up since I wouldn't know what I'd be doing.

BTW, that was a wicked video you've tested with. I immediately wanted to get it then saw the ending it was an MTV vote campaign video? That would be a fantastic movie with those actors.