TMPGEnc 2Pass VBR Confusion

[DJRumpy]

About the same setup as mine. You can encode an SVCD disc in about 20 minutes using CCE. DVD resolutinos will encode at about .9-1.2 times the playback speed. No idea what TMPGenc will do though.

[Anonymoususer]

It took around 4 Hours, instead of 36 Hours [before] on a 50 Minute movie...due to the fact I had set the Priority to HIGH, and had everything checked, Ghost Reduction, Simple Color Correction..etc, made an XVCD at an Average Bitrate of 1996, Minimum of 800, and Max at 2520
[Yes I know that's SVCD Bitrate, but who cares this is an experiment],
used 2 Pass VBR, and set it on Highest Very Slow Option...It came out pretty good, haven't burned it on a VCD yet though...now I will try to make an SVCD and Compare the 2...is there some kind of software with which I can zoom in even more, see the pixelation close up...or is there some option in Windows Media Player...? so far soo good... Will try CCE soon...
BTW just realized that the file came out to be 576MB's even though I choose for it to be 800 MB';s...weird...?

If you guys wanna see the pictures I can post them up for you IF you like, but someone has to explain to me how to post them up...will post more comparison results soon...

[scottb721]

Some VBR questions,
1. Should the minimum be raised so that MAX - Average = Average - MIN ?
(so the average is in the middle of the MIN and MAX bitrates).

2. If I can get 1800 CBR bitrate or more, should I bother using VBR ?

3. When I do VBR, max is 2520, I adjust the average bitrate to make the file fit on 99% of the disc but the MIN bitrate defaults to 300.
Should this min rate be increase to 1600 or more to ensure good picture quality for the low action scene frames or is 300 quite sufficient ?

4. Are there any good guides for TMPG VBR, like what bitrates to set ,especially MIN rate? when to use CBR instead ?

Thankyou.

[txpharoah]

I get faster and better quality results using TMPGEnc Plus in "Automatic VBR (CQ_VBR)" mode than anything else, including other TMPGEnc options and anything from CCE, Cleaner, or MainConcept/Adobe. It's set for 6MPS max, and 2MPS min (no min padding though), with average of 50% to give me about 4MPS average (which would be comparable to 8MPS at 720x480, which many believe to be perfect quality bit/res).

Set your motion precision to "low quality". That does NOT mean you are getting lower quality, it means that the encode for the moving area will have less time spent on it by the encoder. In many cases I find the motion to be smoother, as many encoders have a bad habit of overdoing it, making the motion more jerky with the "higher" quality. Try a panned area of a video and test this on your own. Try low vs. highest, and watch the choppiness appear from the so-called "highest". They really should rename these settings to "Motion Complexity" or something, with one being fastest (low), and one being slowest (not highest). Only occasionally will artifacts appear, but that's on motion so fast, you're not likely to notice. Plus most artifacts are determined by bitrate and resolution anyway.

At 352x480 (true S-video DVD resolution), running deinterlace filters, noise filters, clipping, and simple color corrections, the 2 hours of footage will take 8 hours to encode MPEG2, and quality far surpasses anything I can get anywhere else. System is P4 with RDRAM. For those that don't know, all those filters usually doubles or more the encode time, so it would be a few hours at most for a straight conversion. And don't whine about the hours it may take. Go to bed, done the next morning. You can also do 2-3 during the day, and this quality is better than a video quickie that many amatuers push for. But yes, anything over 8 hours is a bit ridiculous.

Quality is determined by many factors, including your motion precision, GOPs, quantize matrix, etc, not just the bitrates. Time is determined by your RAM and processor, as well as the other specs of your PC, and how many other programs/network connections may be running in the background to leech your resources. (Hint: unplug network cables that open your 65535 ports to open accesses usually unknown to you such as DSL/cable/router pings, and shut down system tray programs like anti-virus that will slurp up your PC power because of how Windows is built.) Also be sure that in TMPGEnc, you set your TASK PRIORITY (in the Options of the toolbar) to maximum. Otherwise, it's not being used to its full potential.

My final product fits on a DVD just fine. I have dozens of them as proof, and I couldn't be happier about the quality. I've been at this for years, and have nothing for contempt for programs such as CCE, Cleaner, and MainConcept that makes programs 1) too expensive, 2) user UNfriendly, 3) difficult to attain good quality, and 4) slowness. TMPGEnc Plus is a pioneer, providing users with an easy interface, but not an amateur product, as it provides quality and timing that can far surpass its supposed higher rivals.

Give my settings a try, and look over your results. These can also be applied to VCD/SVCD/XVCD, just change your bitrates to the well-placed suggestions of 300 min and 2520 max on SVCD, and CBR at 1856 for VCD, experiment on XVCD, and leave GOPs open for CD-based video.

And before anyone slams me out there, as tends to happen on these boards, you'd better be able to put your professional qualifications up with your slam. I work at a place where we make more DVDs in one day that most people on this board will make in a year.

Hope I could help some of you out. It's hard to learn this on your own, passing along some experience.

[DJRumpy]

I'm sorry, but I just can't let this one pass by...

Set your motion precision to "low quality". That does NOT mean you are getting lower quality, it means that the encode for the moving area will have less time spent on it by the encoder. In many cases I find the motion to be smoother, as many encoders have a bad habit of overdoing it, making the motion more jerky with the "higher" quality.

You have some strange notions about MPEG. The motion detection does indeed affect how long the encoder spends on any particular frame, but it directly affects bitrate allocation for a frame, depending on what motion is taking place within a frame. Choosing a setting of lower quality will only net you poorer quality on high motion scenes, because the encoder will not properly detect motion vectors within a frame.

It's set for 6MPS max, and 2MPS min (no min padding though), with average of 50% to give me about 4MPS average (which would be comparable to 8MPS at 720x480, which many believe to be perfect quality bit/res).

1-Pass CQ mode doesn't use an average per say. It simply allocates bitrate to any given frame, as much, or as little, as is needed, to meet the needs of any given frame, up to, but not exceeding your max (or min on the low side). You can indeed get good quality with CQ mode, but predicting size is impossible to any degree of accuracy. This is not a good option if space is at a premium, or the video is contains a lot of motion, as your video can become very large, or you have to reduce your maximum bitrate to make it fit onto your DVD.

Set your motion precision to "low quality". That does NOT mean you are getting lower quality, it means that the encode for the moving area will have less time spent on it by the encoder. In many cases I find the motion to be smoother, as many encoders have a bad habit of overdoing it, making the motion more jerky with the "higher" quality.

If your getting jerky video, it's not due to the motion detection setting. It could be any number of things, including wrong field order, excessive bitrate, buffer underrun, buffer overflow, bad media, bad mpeg, etc.

Only occasionally will artifacts appear, but that's on motion so fast, you're not likely to notice. Plus most artifacts are determined by bitrate and resolution anyway.

If your encoding to DVD, you should never see macroblocks. There is more than enough bitrate to handle input, assuming your source is clean.
The artifacts appear, because the encoder failed to detect where all of the images were moving ( motion vectors) from frame to frame. Instead, it assumes that these moving bits of the image are new images, and it applies bitrate to produce them in that frame. As it fails to detect where these images are moving, due to a low (poor) motion detection setting, more bitrate is applied to produce each frame, until a bitrate shortage occurs. When this happens, you get macroblocks (artifacts), because the encoder does not have enough bitrate to reproduce the frames details accurately.
There's a decent description of it here:
http://nickyguides.digital-digest.com/keyframes.htm
and here
http://www.cs.cf.ac.uk/Dave/Multimedia/node259.html
Basically, if a car drive across the screen, say from left to right, and the encoder doesn't properly detect that the cars motion is taking it across the screen, it would waste bitrate from frame to frame to re-produce the car's image in each frame. A more intensive motion detection setting force the encoder to do an intensive scan of each frame to determine if the image moving from frame to frame is the same object. If it is, it can then supply motion vector information only for the object, and not waste bitrate actually reproducing the full image.

My final product fits on a DVD just fine. I have dozens of them as proof, and I couldn't be happier about the quality. I've been at this for years, and have nothing for contempt for programs such as CCE, Cleaner, and MainConcept that makes programs 1) too expensive, 2) user UNfriendly, 3) difficult to attain good quality, and 4) slowness.

<begin pulpit preaching>
I find it difficult to believe you've ever used CCE, as the output in general matches, or exceeds that of TMPGenc, with no effort invovled. It also does so in 1/10th the time. The cost is the same as TMPGenc for the basic package. I'm sure others here, using the other encoding packages you've mentioned, probably produce excellent results on a regular basis. If they didn't, we wouldn't be using their products. Did you ever stop to consider, that if your results with these products were poor, perhaps you were at fault?
<end pulpit preaching>

[hustbaer]

@txpharoah:

i believe you might have checked the "do not motion search for half pixel".
if so, uncheck it.
then "high quality" motion-search should be as smooth as "low quality".
of-course it will take even longer.

however, with bitrates above 2mbit/s, motion-search is no big concern.
the movie should look "ok" (by which i don't mean "good" or even "perfect") even if motion-compensation is not used at all.

for me it always worked best having checked "use floating point (i)DCT" and having "no motion search for half pixel" unchecked.
and having set motion-search to "highest quality" of course.
"motion-estimate" is fine too, better than "low quality", but not as good as "highest quality" for the most sources.

bye,
--hustbaer

[Anonymoususer]

Well...I made an SVCD with the same settings I mentioned above...and well it's weird...There is definatly a difference...the Mpeg1 is pretty clear, then I made the Mpeg2, and it's even clearer BUT, I seem to notice pixelation in low motion scenes ?...and I realized (if you read the first posts on the first page) someone mentioned that if you put the minimum more then 300 or 500 it can cause problems...and I realized that I had kinda screwed up...but there is definatly a very huge difference the Mpeg2 seems very Clear and the Mpeg1 is a little grainy, but you can definatly tell the difference...sooo now I will encode again using 2 Pass VBR for TMPGEnc, but this time the Minimum Bitrate will be 300KB/s...will post back soon, if anyone cares...

[scottb721]

I encoded the same AVI clip into mpeg1 and mpeg2.
The mpeg1 had no macroblocks but was grainier and lhad ess detail due to the lower resolution when compared to the mpeg2 version, which was clearer but suffered blocking on fast moving objects. That was even with "Motion seach slow" .

[Anonymoususer]

No the thing with mine is that Macroblocks appear on low motion scenes, as mentioned before on Page 1 someone explained to me that Min bitrate raising can cause problem, and that's what happened...so what did you go with Mpeg1 or Mpeg2...?

[scottb721]

Hi DivXerouS
You'd think a low min bitrate would cause more problems than a higher min rate. Oh well, hopefully someone can explain this someday.
I've been tempted to raise the min rate in VBR but have always just left it at 300.
If I can get a CBR around 1800 or more I tend to just go with that.
I thought this would ensure that no frames were starved of bitrate.

I do all my stuff in MPEG2. Most of my sources are DVD so I think the very high quality of the source helps reduce the incidence of the blocking.
My example I refered to was converting analog camcoder to (S)VCD.

Perhaps the lower quality source is what is causing problems with the quality of the MPEG2 output. Don't get me wrong. The resultant SVCD is quite acceptable, just extra blocking than I get with DVD's transfer to SVCD.
Another reason I like SVCD is getting to do chapters and playback control using VCDEasy. (I'm not sure of VCD's ability to have these features)

Cheers

[Anonymoususer]

Yes I do prefer SVCD's but I am kinda of testing, cuz I just got a new system, you know what I mean...

If you wanna read about more on that try reading the Post

2 Pass VBR Confusion...hope it helps..

[txpharoah]

I'm sorry, but I just can't let this one pass by...

Ditto...

Set your motion precision to "low quality". That does NOT mean you are getting lower quality, it means that the encode for the moving area will have less time spent on it by the encoder. In many cases I find the motion to be smoother, as many encoders have a bad habit of overdoing it, making the motion more jerky with the "higher" quality.

You have some strange notions about MPEG. The motion detection does indeed affect how long the encoder spends on any particular frame, but it directly affects bitrate allocation for a frame, depending on what motion is taking place within a frame. Choosing a setting of lower quality will only net you poorer quality on high motion scenes, because the encoder will not properly detect motion vectors within a frame.

Consider this: this is just time spent, it doesn't mean the time is spent well. Let's say I was a student. Now let's say I was taking a test. And question #1 question is "Who is the 1st President of the U.S.?" Now, I could spend 10 seconds or 10 minutes on this question, but would most likely end up with the same answer (Wasington) at the end of both periods. Same can be said here. The extra time CAN be more useful but I find it to NOT be. In fact, I find the results of spending too much time may very well unnnecessarily underflow the buffer for no reason. Even without jerky motion on Highest, the encode quality still looks as good as Lowest or Estimate. May be an error with TMPGenc on any MPEG2 encoding. I would have to suggest Highest on CBR MPEG1 encoding.

It's set for 6MPS max, and 2MPS min (no min padding though), with average of 50% to give me about 4MPS average (which would be comparable to 8MPS at 720x480, which many believe to be perfect quality bit/res).

1-Pass CQ mode doesn't use an average per say. It simply allocates bitrate to any given frame, as much, or as little, as is needed, to meet the needs of any given frame, up to, but not exceeding your max (or min on the low side). You can indeed get good quality with CQ mode, but predicting size is impossible to any degree of accuracy. This is not a good option if space is at a premium, or the video is contains a lot of motion, as your video can become very large, or you have to reduce your maximum bitrate to make it fit onto your DVD.

I'm using DVD. Space means nothing to me. I can fit 3 hours movies on a disc just fine or several episodes of a tv show in immaculate quality. Run out of space? Get another disc.

Set your motion precision to "low quality". That does NOT mean you are getting lower quality, it means that the encode for the moving area will have less time spent on it by the encoder. In many cases I find the motion to be smoother, as many encoders have a bad habit of overdoing it, making the motion more jerky with the "higher" quality.

If your getting jerky video, it's not due to the motion detection setting. It could be any number of things, including wrong field order, excessive bitrate, buffer underrun, buffer overflow, bad media, bad mpeg, etc.

None of this.
Bad media? It's not even off the system yet.
Jerkiness can be caused by stupidity from an over-thinking motion search or by GOP errors. You may be on to something with the buffer. But even set automatic, why would the program crash itself by one of its own settings?

Only occasionally will artifacts appear, but that's on motion so fast, you're not likely to notice. Plus most artifacts are determined by bitrate and resolution anyway.

If your encoding to DVD, you should never see macroblocks. There is more than enough bitrate to handle input, assuming your source is clean. The artifacts appear, because the encoder failed to detect where all of the images were moving (motion vectors) from frame to frame. Instead, it assumes that these moving bits of the image are new images, and it applies bitrate to produce them in that frame. As it fails to detect where these images are moving, due to a low (poor) motion detection setting, more bitrate is applied to produce each frame, until a bitrate shortage occurs. When this happens, you get macroblocks (artifacts), because the encoder does not have enough bitrate to reproduce the frames details accurately. There's a decent description of it here: http://nickyguides.digital-digest.com/keyframes.htm
and here http://www.cs.cf.ac.uk/Dave/Multimedia/node259.html

Never say never. Any encode can have noise, and my whole point was that even if any occurred, they would be small and unnoticeable. Even 10MPS ecnodes can have noise on highest settings. It happens. Digital video is not perfect yet.

Basically, if a car drive across the screen, say from left to right, and the encoder doesn't properly detect that the cars motion is taking it across the screen, it would waste bitrate from frame to frame to re-produce the car's image in each frame. A more intensive motion detection setting force the encoder to do an intensive scan of each frame to determine if the image moving from frame to frame is the same object. If it is, it can then supply motion vector information only for the object, and not waste bitrate actually reproducing the full image.

Yes, on high motion, you may need to consider alternative motion settings, but then again, I'd suggest hardware motion detection if you're needing such things. Most movies have no high motions as you've described.

My final product fits on a DVD just fine. I have dozens of them as proof, and I couldn't be happier about the quality. I've been at this for years, and have nothing for contempt for programs such as CCE, Cleaner, and MainConcept that makes programs 1) too expensive, 2) user UNfriendly, 3) difficult to attain good quality, and 4) slowness.

<begin pulpit preaching>
I find it difficult to believe you've ever used CCE, as the output in general matches, or exceeds that of TMPGenc, with no effort invovled. It also does so in 1/10th the time. The cost is the same as TMPGenc for the basic package. I'm sure others here, using the other encoding packages you've mentioned, probably produce excellent results on a regular basis. If they didn't, we wouldn't be using their products. Did you ever stop to consider, that if your results with these products were poor, perhaps you were at fault?
<end pulpit preaching>

Come down from the pulpit preacher. I've used them all. They all have their uses, and they all have made excellent products for me. But in terms of ease of use, in terms of speed, and in terms of preview capabilities, I support TMPGenc PLUS far more than any other encoders. I bought them all. I own them all. I have novels (manuals) for them all.

Customers flat out complained about the noise from CCE. Period. No way around that. I called, I tried, I looked: it's just how the thing works. Cleaner has piss-poor customer service. It likes to crash, and has had more updates and patches than I can count. MainConcept (Adobe) is difficult to use and understand, but I use it when I have a Premiere project. Its quality is as good as TMPGenc, but can have a blurry encode sometime. The MainConcept understanding is surely my failing (as well as a badly-written manual and no help from Adobe), but the speed is not.

I'd buy bigger, meaner hardware setups, but have no reasons as of yet, nor does the place where I work.

My suggestion to the hobbiest is surely TMPGenc in terms of costs and ease of use. I give out settings that will work for most of the people most of the time. If you are not a hobbiest, why are you here? I come to help and give advice to newbies, in addition to see what's new in the world of video.

[DJRumpy]

The extra time CAN be more useful but I find it to NOT be. In fact, I find the results of spending too much time may very well unnnecessarily underflow the buffer for no reason.

It sounds like your confusing playback with encoding. Once the encoder encodes the mpeg, it's all just bitrate after that. The VBV circuit would throttle back the mpeg stream in-flow to compensate for bitrate spikes, and allow the buffer to fill on low bitrate scenes, maintaining a constant output. If your worried about buffer underflow, have you ever considered using the settings that are there for that purpose? Try using padding, rather than making your high motion scenes suffer. It's designed specifically to prevent your bitrate from dropping to low.

I'm using DVD. Space means nothing to me. I can fit 3 hours movies on a disc just fine or several episodes of a tv show in immaculate quality. Run out of space? Get another disc.

Even a newb quickly finds out that space does mean something, even on DVD, otherwise we'd all be happily copying our dual layer DVD's, and not even bother to learn how to re-encode to begin with.

Bad media? It's not even off the system yet.
Jerkiness can be caused by stupidity from an over-thinking motion search or by GOP errors. You may be on to something with the buffer. But even set automatic, why would the program crash itself by one of its own settings?

You didn't specify it was jerky on your PC, or that you had crashing problems. If the video is jerky, before you even burn it, then of course it wouldn't be media. Perhaps you should be more specific when making such a statement. You also claim the program is 'stupid' using it's programmed motion search techniques, and then ask why would the program crash itself? (not agreeing here, as I think TMPGenc happens to be an excellent piece of software. Just pointing out inconsistancies in your post). Again, I have to ask, have you ever considered that perhaps your doing something wrong? Judging from your response, and the attitude conveyed in your words, I'd have to guess no.

Never say never. Any encode can have noise

You are absolutely correct. I have said the same thing many times. I should know better.

Come down from the pulpit preacher. I've used them all. They all have their uses, and they all have made excellent products for me. But in terms of ease of use, in terms of speed, and in terms of preview capabilities, I support TMPGenc PLUS far more than any other encoders. I bought them all. I own them all. I have novels (manuals) for them all.

Strange that you have gotten excellent results from all of them, yet you have nothing but contempt for them. You should at least try to keep your 'contempt' level consistent for arguments sake.

I've been at this for years, and have nothing for contempt for programs such as CCE, Cleaner, and MainConcept

I have been take to account many times in this forum. It's a learning process. Laying blame on a piece of software, without even considering other options is foolish. A more constructive method might have been to suggest that the setting caused the problems your seeing, and asking someone else to verify your results.

I too have been at this for years. I'm still learning...