2] How many avi without losing quality?

Avi to DVD and other video conversions, such as wmv to DVD, mts to DVD, mkv to DVD and more. 2 pass encoding, multi-core support, and always more files supported.

Moderators: Maggie, JJ, ckhouston, Phil, Forum admin

User avatar
Coral
VSO Moderators
Posts: 1526
Joined: Fri Feb 06, 2009 7:10 am
Location: Malta

Post by Coral » Sun Mar 22, 2009 10:22 am

I compleatly agree with ckhouston. You will find that with the misnomer "Encodeing Quality" the difference in quality between High and Medium is most of the time insignificant. If you use qood quality material the end result is also good.

If you want to "squeeze" in a large project, use dual layer or if the VIDEO_TS folder contents are too large for a single layer but much smaller for a dual layer use a dvd squeezer. I use DVD2one and the result is still quite good.

Last but not least. The hardware you play your final product is also important. Not all home dvd players decode to the same quality and some graphic cards are to slow to handle fast decoding. I have seen movies that looked horrible on a PC but quite bearable on a home dvd player.

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Wed Apr 08, 2009 1:09 am

Thanks for the comments. I think it is important to clarify this thought though.
You will find that with the misnomer "Encodeing Quality" the difference in quality between High and Medium is most of the time insignificant.
There is little difference in converted quality between those two settings only when High does not fill the target. When High fills the target, Medium will usually produce better results, sometimes much better.

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Wed Apr 08, 2009 2:18 am

I continue to see posts anticipating the upcoming 2 pass conversion for CX2D. I look forward to it also but I don't expect it to provide much improvement over the present 1 pass conversion when the proper choice of the so called "quality" setting is used. I am not alone, VSO staff have aready said in this forum that they cannot find much improvement with the 2 pass, or something to that effect.

I know that there will be those that believe the 2 pass is significantly better regardless of any facts that show it isn't, just like there are those that refuse to believe that High is only a choice of an encoding profile that does not always produce better results than Medium, and that Medium is usually the better choice for most movies. But for those willing to be objective, I want to explain why no encoder can do much better than the present 1 pass as long as the proper setting is used.

Look at the bitrate plots on the previous page, they show how bitrate is distributed over a two minute time period of the video. The best encoder will distribute bitrate as needed for the different scenes in that time period as efficiently as possible within the constraints of target size and DVD standards.

Because the source came from a DVD distributed by a major movie studio, we can assume that the encoder used for it was the best money can buy and that considerable time was spent by very qualified professionals in encoding it while using expensive hardware. In other words, the shape of that source plot is probably the best that money can buy and is therefore the ideal that any encoder should try to reproduce.

Now compare the shape of the source plot with those produced by VSO conversions. Medium and even Low produce shapes that mimic that of the source pretty well as long as they do not fill the target and need additional compression.

The source was a movie about 145 minutes long as I remember. I did not provide plots for conversions of the same length, but I can say emphatically that, if the original movie was converted instead of using my test files, one of the settings, either Medium or Low, would mimic the shape of the source very well.

Since the ideal is to reproduce the source shape as closely as possible, what more can we expect from any encoder than what we already have with the present 1 pass encode? It really can't be improved much if we use the proper setting choice. A 2 pass may refine the shape some, but there is little chance it can provide noticeably better results except maybe for projects longer than typical movie lengths.

JoeB
VSO Fanatic
Posts: 4348
Joined: Mon Jul 02, 2007 7:52 pm
Location: Canada

Post by JoeB » Wed Apr 08, 2009 2:47 am

ckhouston wrote: [Lots of stuff snipped for brevity]

Since the ideal is to reproduce the source shape as closely as possible, what more can we expect from any encoder than what we already have with the present 1 pass encode? It really can't be improved much if we use the proper setting choice. A 2 pass may refine the shape some, but there is little chance it can provide noticeably better results except maybe for projects longer than typical movie lengths.
Over time, ckhouston, people may - and many likely will - get the idea. That pre-supposes that they read the stuff and - if unconvinced - go back to some of your ealier posts that were very detailed in terms of graphs, explanations, etc.

I just hope that VSO doesn't spend too much time working on this 2-pass thing (which, as you already know, won't make a noticeable difference in the viewed output) at least until they have added a menu editor (which is something that you can use and see the difference with :-)

Regards,

JoeB

jolo
Posts: 7
Joined: Sun May 03, 2009 6:05 am

Isn't movement as or more important than scenary ??

Post by jolo » Sun May 03, 2009 7:06 am

ckhouston wrote:How to Determine the Best “Quality” Setting to Use - SUMMARY

For projects converted to a DVD-5 (4300 MB) Target size at Full D1 DVD resolution, the best simple rule of thumb I can give for videos with typical scenery is:

1. Use the High setting for projects less than 80 minutes long.
2. Use the Medium setting for those between 80 and 160 minutes long.
3. Use the Low setting for those over 160 minutes long.

160 x (4300/4300) x 1.25 = 200 min

So you should use the Medium setting since your total project length of 120 min falls between 100 and 200 min.
Excellent post.

I have encoded video for many years, in the days where MPEG-1 was about as good as one could get to play on a DVD player.

It has been said several times and can't be said enough that the most important factor in getting quality output is the quality of the source video.

Another strong factor is what device is being used to watch and listen to the video.
In general, when encoding, watching a video or audio, the quality will drop to the weakest link in the chain. Like listening to song using a (ugh) Apple Ipod

In the "older" days, I used to encode a small sample to see how to configure t he output settings.
The sample that I would look for always would contain dialog and most important, contain a lot of movement.

This was implied earlier in this discussion when it was mentioned that a soccer game would take more resources than a National Geographic HD type of video.
I certainly agree with that, for everything that I have learned about Video, but perhaps VSO has some other method when encoding.

Encoding a video that has a high level of fast movement were the ones that seemed to be the largest struggle because of the resources needed to encode the output.

So far, this discussion only seems to focus on scenery. Heck, if a video was just scenery, take a look at the size of image files as compared to video files, even if they contain a high quality audio.

The last thing that Variable Rate encoding is great, especially if there is a lot of movement, as more pixels could be given to those action scenes and taken from scenes that have little or no movement.

I do lots of encoding to Divx, and share video with others, like myself, who have a stand lone DVD player that is Divx certified. For those who don't have that, then DVD format is what I might need.

Of course length of the video is a major concern with quality, but I cannot understand how ConvertXtoDVD could make a 20 minute video look better using medium than high. It would seem to me that both high and medium might look the same because a blank DVD has more than it needs to handle the best quality that can be created. I would expect that High might use more disk space, and increase the time needed to encode, but not increase the quality if there is only 20 minutes used.

Thanks,

Jon :D
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Blu-Ray is for suckers !!

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Mon May 04, 2009 1:07 pm

Thanks for your comments.

The purpose of these posts is to convince users that the so called "quality" settings are not what they seem. It has not been easy so I have had to go into more detail than most want to see because video encoding is a complicated subject. I can't keep repeating everything in each post so it unfortunately is necessary to read several long threads to completely understand my point(s). And it is necessary to pay attention to wording instead of filtering out words and concepts we are already familiar with and tend to agree with, we are all inclined to do that because it is easier and saves time, but it can give a wrong impression.

Not only have I had to surmount natural skepticism that what I say is correct in view of wording used by VSO to describe the "quality" settings, I have had to counter popular myths such as bigger output means better quality and that video action determines primarliy how much bitrate is needed. These myths probably started because of our desire to reduce every subject to one word or sentence no matter how complicated the subject. It just does not work and causes a lot of cunfusion and misdirected efforts.

It has been a long and tiring task so I may get overly defensive when someone questions what I have said without seeming to understand. So I'm going to do some nitpicking with your post in order to prevent confusion, nothing personal.
jolo wrote:Of course length of the video is a major concern with quality, but I cannot understand how ConvertXtoDVD could make a 20 minute video look better using medium than high. It would seem to me that both high and medium might look the same because a blank DVD has more than it needs to handle the best quality that can be created. I would expect that High might use more disk space, and increase the time needed to encode, but not increase the quality if there is only 20 minutes used.
I didn't say that, I did say
High should give better quality for that 20 min video because there will be plenty of resources available to waste on simple scenes, but you may not notice any improvement over Medium even then except a slight difference in simple scenes.
jolo wrote:Encoding a video that has a high level of fast movement were the ones that seemed to be the largest struggle because of the resources needed to encode the output.

So far, this discussion only seems to focus on scenery. Heck, if a video was just scenery, take a look at the size of image files as compared to video files, even if they contain a high quality audio.
You used the word movement at least, but many will take what you said as action is the determining factor so action movies must put the biggest strain on resources and dismiss eveything else as BS.

Here is what I said in this thread
Encoding is a matter of compression and how much scenery can be compressed depends on the scenery itself. Look at the plots above. Notice the bitrate for some scenes (simple scenery) is low while that for others (complex scenery) is much higher. Simple scenes can be compressed a lot, complex scenes cannot, they require more data (bitrate) to describe them as accurately as simple scenes.

Complex scenes have a lot of detail like many sharply defined contrasting edges, many areas with varying visual properties like color or brightness, etc., and/or motion including camera pans and zooms. Simple scenes are the opposite. They include dark scenes, most closeups and even action scenes where the background is blurred due to camera pans.
The word scenes is used in encoding to mean a series of frames, usually several seconds long, with some commonality, not a single frame. That is what I mean with terms like complex scene/scenery. I tried using the more general term of complex video content at one time but that didn't seem to convey as much meaning as complex scenery.

Compression is basically done by lumping pixels together that have similar visual properties so they can be described as one. Then motion is accounted for by accounting only for differences in the next frame and so on. An occasional complete refresh is done like in the first frame to keep things accurate.

Much resources have to be used for the first frame if it has a lot of detail to describe, many sharp edges defined by contrasting areas and big variations of visual properties within those areas. The next frame will have a more differences to account for if movement in the scene is fast, those differences are also increased though by the high detail.

So compression is determined by the balance of detail and movement and both have to be considered. There is always some movement in a video. Some with a lot of fast moving scenes may not require as much resources as one with slower movement because of differences in background detail. You basically have to think like an encoder to evaluate them or use a program that shows how much resources are appropriated to different scene types. And you have to consider the ratio of simple to complex scenery. A soccer video that spends a lot of time showing wide angle shots of field action does not require as much resources as one that has more closeups of cheering crowds.

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Thu Jun 25, 2009 5:47 am

I have noticed that a lot of users still do not follow my recommendations about how to choose the so called “quality” settings to get the best quality converted results. I have basically said that one must use those settings to prevent the conversion from filling the DVD in order to get the best quality conversion.

I have presented data to support that contention but it seems to be ignored by many, maybe because data is boring, but there are other possible reasons. The recommendations are contrary to implied meanings of the setting names and descriptions, and VSO has not yet stepped in to clarify the issue. And conventional wisdom is that bigger conversions that fill the disc are better. That is true for traditional variable bitrate (VBR) encoders but not for the unique one used by CX2D - more about this later.

The boring data that has been presented may become more interesting and, hopefully, more believable, if one understands the encoder and why it produces conversions with those characteristics. I will try to keep this as concise as possible but there will be many that stop reading when they see its length. That is up to you, but I will say that you are missing much of the potential of CX2D if you do not use my setting recommendations.

There is a good discussion of encoding methods in http://tangentsoft.net/video/mpeg/enc-modes.html that may help clarify some discussion below.


Understanding the Uniqueness of the CX2D Encoder

Most people are familiar with constant bitrate (CBR) encodes and traditional variable bitrate (VBR) encodes. CX2D uses a modified constant quantization factor (CQ) encode that many aren’t familiar with.

A simplified explanation of quantization factors (Q) is given in the post at the bottom of this page http://forums.vso-software.fr/bad-algor ... t4378.html Note that lower Q represents higher quality, higher Q is lower quality.

A pure CQ encode holds Q constant and allows bitrate to swing up and down as much as needed to keep Q constant. It will yield results superior to even multipass traditional VBR, if Q is low enough (Q is an integer and the lowest value is 1). The reason pure CQ is not used more is because the converted output size cannot be predicted with any degree of certainty - the bitrate distribution depends very much on the scenery in a video, so it can be very different for different source files and there is no way to know until the conversion is actually done. Output size of CBR is easily predicted. And traditional VBR constrains bitrate with either an average value or by using upper and lower limits - this is why bigger is better with them BTW - so output size is reasonably predictable.

CX2D does constrain Q with the “quality” settings to 1 - 31 for High, 2 - 31 for Medium and 3 - 31 for Low.

The Q plots in the last link above show that the CX2D encode is essentially CQ - at the minimum Q set by the "quality" setting - as long as the target DVD is not filled. There are small peaks at points where the bitrate has to be lowered from the desired value down to the maximum permitted on DVDs, but the plots are nearly constant. I don’t think any traditional multipass VBR will beat CX2D’s result in this case where the disc is not filled. Nor should a multipass CX2D encode based on the same Q model make any difference - Q cannot be made anymore constant than it is with 1 pass.

But things change when the target disc is filled. Bitrate has to be constrained then, maybe by using an average, in order to fit the conversion to a given size. Nearly CQ can no longer be maintained and quality drops. I don’t know exactly what the program does in this range, but it appears the encode essentially transitions from near CQ to a 1 pass traditional VBR, probably using an average bitrate, for long enough projects.

Hopefully, the need to use my setting recommendations is more clear now. One has to keep the conversion from filling the target in order to get the best result. And my recommendations based on project length are designed to do just that for the majority of video that most users will convert.


Update: This post was written before version 4 was available for testing. The plots in a post below confirm my conclusion above that 2 pass would not make a difference if a conversion does not fill the DVD, because that conversion is essentially CQ and Q cannot be made any more constant than it is with 1 pass.

The new plots also seem to confirm my assumption above that a conversion that does not fill the DVD behaves more like traditional VBR than CQ. 2 pass does make a difference with traditional VBR and the plots do show significant improvement for conversions that do fiil the DVD.
Last edited by ckhouston on Tue Nov 03, 2009 12:12 pm, edited 2 times in total.

erk
Posts: 53
Joined: Fri Dec 28, 2007 9:24 am
Location: Washington, DC
Contact:

Re: Isn't movement as or more important than scenary ??

Post by erk » Sun Jul 05, 2009 6:49 am

ckhouston wrote:How to Determine the Best “Quality” Setting to Use - SUMMARY

For projects converted to a DVD-5 (4300 MB) Target size at Full D1 DVD resolution, the best simple rule of thumb I can give for videos with typical scenery is:

1. Use the High setting for projects less than 80 minutes long.
2. Use the Medium setting for those between 80 and 160 minutes long.
3. Use the Low setting for those over 160 minutes long.

160 x (4300/4300) x 1.25 = 200 min

So you should use the Medium setting since your total project length of 120 min falls between 100 and 200 min.
This is as basic as it gets...there shouldnt be anymore questions about it really "what quality setting i should use"

Just look at what he said...pretty simple..

Im going to start using these settings right away...I even saved this post as a text doc. inside my movies folder for quick reference in case i forget.

Thanks MAN !!! big props.

im mad i been using HIGH setting on all my movies in the past..i wish i knew this even before. but oh well.

EDIT: Can some 1 clear this up for me 1 more time, I got confused on the last line this guy mentioned..and i noticed the software doesn't say Hr it always says min.

If the file says "Titleset ~120min." That means its actually 80min which = 1hr 20min Correct ? (Which means you should use the Med. Quality setting for this example length)

cuz the program is sayin minutes instead of hours. And that kind of through me off a bit. I mean i still know how many min. is in a Hr. But this might confuse people, if it hasn't already.

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Re: Isn't movement as or more important than scenary ??

Post by ckhouston » Mon Jul 13, 2009 2:34 pm

erk wrote:EDIT: Can some 1 clear this up for me 1 more time, I got confused on the last line this guy mentioned..and i noticed the software doesn't say Hr it always says min.

If the file says "Titleset ~120min." That means its actually 80min which = 1hr 20min Correct ? (Which means you should use the Med. Quality setting for this example length)

cuz the program is sayin minutes instead of hours. And that kind of through me off a bit. I mean i still know how many min. is in a Hr. But this might confuse people, if it hasn't already.
I'm not sure what last line confused you.

Titleset times are shown in minutes. So 120 min = 2 hr, not 80 min = 1 hr 20 min as you show.

Total project length (the sum of all titlesets as shown in the CX2D status bar) is the time you need to be concerned with. It is is given in hr:min format so it has to be converted to minutes to use the setting guidelines. Multiply the hrs by 60 and then add the minutes to get total minutes.

Let me know if it is still not clear.

User avatar
Coral
VSO Moderators
Posts: 1526
Joined: Fri Feb 06, 2009 7:10 am
Location: Malta

Post by Coral » Mon Jul 13, 2009 3:08 pm

Graphically it would look like this:
Attachments
Titleset.jpg
Titleset.jpg (72.47 KiB) Viewed 9271 times

mohammad wong
Posts: 5
Joined: Thu Jul 16, 2009 7:07 pm

Post by mohammad wong » Wed Jul 29, 2009 6:54 pm

So, to rewrite the rule of thumb: (Bold/Red are my changes)
ckhouston wrote:How to Determine the Best “Quality” Setting to Use - SUMMARY

For projects converted to a DVD-5 (4300 MB) Target size at Full D1 DVD resolution, the best simple rule of thumb I can give for videos with typical scenery is:

1. Use the High setting for projects less than 01:20:00 long.
2. Use the Medium setting for those between 01:20:00 and 02:40:00 long.
3. Use the Low setting for those over 02:40:00 long.

where the project length is the Total time shown in the CX2D status bar as source files are added to the project. (The status bar is at the bottom of the main CX2D window.)

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Fri Oct 02, 2009 8:31 am

2 Pass Conversions with Version 4

Tests were done to compare 1 and 2 pass conversions in version 4. The same 20 minute test file used for other tests discussed in this thread was used. It was repeatedly added as many times as needed to build projects of different lengths, so the same scenery was converted in all tests. Conversions were made for DVD5 discs using Full D1 resolution.

Plots of bitrate (BR - yellow) and quantization factor (Q - green) are presented below for the same 2 minute segment of 60 to 180 minute project conversions. Vertical scale is 0 to 10,000 kb/s for BR and 0 to 10 for Q. (Note: If you have trouble distinguishing the BR and Q curves, right click on the page and choose to Refresh Page with Full Quality)

Shorthand notations are used as follows: SP designates the Short project encoding option (same as High in ver 3) MP for Medium projects (Medium in ver 3) and LP for Long projects (Low in ver 3). 1P and 2P means 1 pass and 2 pass encodes.

The first plot shown for each combination of encoding option and number of passes is the last one before the DVD is filled. All plots to the left of them would be identical, all to the right are not because they filled the DVD. I evidentally messed up the conversions for MP at 160 and/or 180 minutes, but they do not affect the discussion or conclusions below.

Discussion

Compare the 1P and 2P plots for conversions that did not fill the DVD. They appear identical but there are actually some small differences where Q peaks occur due to BR being constrained to DVD max allowed BR. The differences are not discernable during playback though, so 2P offers no advantage for conversions that do not fill the DVD. They are a waste of time in that case.

2P does improve the conversion though when the DVD is filled. Compare SP 1P, SP 2P and MP 1P at 100 min. MP 1P shows that the different scenes are about 7 to 20 seconds long and shows BR peaks of the same duration for complex scenes and dips for the simple scenes. SP 1P does not cover those scenes as well, it shows sharp momentary peaks that do not cover the entire scene instead. The reason is that the encoder sees a complex scene and starts to apply the appropriate higher BR, but realizes it has not saved enough resources to keep on, so it suddenly drops the BR. Now notice that SP 2P does define the scenes much better than SP 1P, because it has enough information to know it should conserve resources in simple scenes in order to provide more for the complex ones. So 2P does have an advantage when the DVD is filled.

The question then becomes which is better at 100 min, SP 2P or MP 1P? It really depends on personal preferences and the type of scenery in the video. SP 2P will give better quality for the simple scenes but less for complex ones. For typical video, MP 1P usually gives the best result, but SP 2P will improve dark scenes if they are an important part of the video. The choice is easier at 120 min, MP 1P is the clear winner there as indicated by both BR and Q plots.

Conclusions

I personally do not find much benefit for using 2P except for a couple of cases. If pixelation in dark scenes is a problem for projects less than about 110 minutes, try SP 2P. And for projects longer than about 160 min (Edit Note: Sorry that should be 170 or maybe 180 min), use LP 2P if you want the best possible and are willing to allow the increased conversion time for 2P.
Attachments
1 and 2 pass comparison.jpg
1 and 2 pass comparison.jpg (230.54 KiB) Viewed 8750 times
Last edited by ckhouston on Thu Feb 04, 2010 6:57 am, edited 2 times in total.

erk
Posts: 53
Joined: Fri Dec 28, 2007 9:24 am
Location: Washington, DC
Contact:

Post by erk » Mon Nov 16, 2009 4:55 pm

mohammad wong wrote:So, to rewrite the rule of thumb: (Bold/Red are my changes)
ckhouston wrote:How to Determine the Best “Quality” Setting to Use - SUMMARY

For projects converted to a DVD-5 (4300 MB) Target size at Full D1 DVD resolution, the best simple rule of thumb I can give for videos with typical scenery is:

1. Use the High setting for projects less than 01:20:00 long.
2. Use the Medium setting for those between 01:20:00 and 02:40:00 long.
3. Use the Low setting for those over 02:40:00 long.

where the project length is the Total time shown in the CX2D status bar as source files are added to the project. (The status bar is at the bottom of the main CX2D window.)
thanks !!! this cleared it up completely for me.

Lordism
Posts: 1
Joined: Mon Nov 23, 2009 6:55 pm

Post by Lordism » Mon Nov 23, 2009 8:04 pm

Thanks, ckhouston...great info!

ysm79
Posts: 7
Joined: Mon Jul 26, 2010 12:18 pm

Re: 2] How many avi without losing quality?

Post by ysm79 » Sat Sep 11, 2010 9:27 pm

jjfan19 wrote:how many xvid videos can i put onto the dvd without losing the quality of the source avi?

for instance i have 2 avi's (1200 and 1100 kbps), would encoding them in high quality setting be better or same as medium quality? from the vso board i see a sticky that mentions that medium quality is fine, and should be used with any project over 1hr20min?

now i tested both high and med settings.. high gives me average bitrate of 3600kbps .. whereas medium gives 2500kbps. yet when i play it back on standalone dvd player, i notice no difference really.. maybe medium looks better, does this make sense?

basically should i be going with whatever the bitrate on the avi is? i mean most avi are 700-1100kbps range.. so sounds like i can put more then 3 of them on 1 dvd without losing any quality from it at all.. true?
Where is the High quality settings? Is that the Encoding options?

ckhouston
VSO Fanatic
Posts: 4590
Joined: Wed Aug 08, 2007 4:12 am

Post by ckhouston » Thu Oct 14, 2010 3:01 pm

The method of encoding with the different options is the same in ver 4 as in earlier versions. Only the names/descriptions are changed in ver 4 to better explain how the options should be chosen:

ver 3 High Quality = ver 4 SP (Short Projects) --> Up to 80 min.

ver 3 Medium Quality = ver4 MP (Medium Projects) --> Up to 160 min.

ver 3 Low Quality = ver 4 LP (Long Projects) --> 160 min or more.

Version 4 also adds an Automatic option that chooses the best encoding option for projects of different lengths. That had to be done manually in earlier versions.

Post Reply