CBR

[cartman0208]

... Even if the CBR and VBR streams were both identical bitrates the CBR one would be better in aggregate over the course of the movie because the actual encoding settings are better. So yes, while VBR might have occasional massive bitrate spikes that make those brief seconds better quality, it's also has bitrate drops, and the encode settings are worse, so overall CBR is better for 99% of the movie....

I could do that with a whole movie .. and depending on the content, VBR would still look better. Unfortunately I can't post the result here, that would be against forum rules

What better encoding settings do you mean? I pointed out, that the settings were exactly the same.

And despite repeating myself: of course you can pump up the encoding complexity to increase CBR Quality, but is it worth the time and CPU power?
I don't think Amazon "gimps" the videos just to annoy customers ... even a huge provider has to watch the bandwith, when millions of people are watching. IMO that's the reason not everyone is getting the high bitrate stream.
Additionally that ones are probably better protected than the lower streams, preventing AnyStream to download. (just guessing here)

 

[XeXoX]

AMZ store CBR for everything as far as I know. It's nothing to do with screen recording or any of the sort. Certain devices or browsers get CBR and others get VBR - 9 times out of 10 CBR on AMZ is superior to their VBR streams. But if you get CBR you don't get Dolby Digital Audio but AS can easily implement the video stream from the CBR manifest paired with the VBR manifest to get the audio stream. It's as simple as mixing two different mpd files (they can get them both served using what they're using).

In regards to why AMZ VBR is so poor, it's a matter of fact that they changed their encoding settings about 2-3 years ago on the VBR streams and they no longer look as good as the available CBR streams.

I mistakenly linked to an image hosting website earlier so my apologies. But if you use Fiddler and change the settings similar as per the attached image you should get the CBR stream (you will need to turn this on and off so you can get the CBR video and then the DDP audio from the VBR manifest)

EDIT: Because 0x0x0x0x0 is such a prick I've removed the relevant tweaks to get the CBR manifest.

FML. I wish I had seen this thread sooner. I have been looking to achieve CBR for ages. Would you please reshare the image again?

 

[EvilMonkey]

I don't really want to wade into this, but there's a lot of misinformation here. The reason CBR is being championed over VBR has less to do with bitrates and everything to do with Amazon's own encode settings. This is the reason you see people refer to Amazon VBR 1.0 vs VBR 2.0 encode settings profiles, they intentionally made VBR worse in 2.0 and left CBR intact. No comparison you do yourself can possible be accurate because the crux of the issue is that Amazon gimps their own VBR stream with worse x264 settings, whereas the CBR stream is encoded mostly correctly. Even if the CBR and VBR streams were both identical bitrates the CBR one would be better in aggregate over the course of the movie because the actual encoding settings are better. So yes, while VBR might have occasional massive bitrate spikes that make those brief seconds better quality, it's also has bitrate drops, and the encode settings are worse, so overall CBR is better for 99% of the movie.

This has been been demonstrated hundreds if not thousands of times over in movie after movie on sites that aren't this one and which aren't allowed to be mentioned here. The AnyStream developers are of course welcome to not add CBR support, it's their product after all, but random users arguing that VBR is better because "reasons" without actually understanding what's going on behind the scenes is counterproductive and adds nothing to the discussion.

That's my two cents, and I'll stay out of it now. The devs know the difference I'm sure, and they can decide what's best for them and for their product.

I thought that I made it clear early on that I was referring to Amazon CBR and not just CBR in general. I stopped posting the other day because I could not get the troll to understand what we were talking about. Now he thinks that Amazon "intentionally" made the vbr stream worse.

Speaking of intentionally making a stream worse, when the pandemic hit Netflix deliberately lower the quality of their streams because of how many people were now at home streaming. I don't think they ever brought the quality back up to where it used to be. So I guess that is one example of a provider intentionally making their quality worse. However, I don't believe that was Amazon's intention but nevertheless their new encoding method still proved to be inferior to the CBR stream.

I suspect the troll will end up making more videos comparing CBR to VBR when the conversation was solely about Amazon.

If he doesn't want to believe what we are saying then that is his choice. At the end of the day his belief is in the minority. Kind of like there is a minority of people that believe Hollywood is full of vampires searching for the blood of children.

 

[jaqob]

It is a pretty well-documented fact in the p2p community that Amazon switched from a good VBR profile to worse profiles in the last few years, making CBR the highest quality stream. The files exist in the p2p community to prove this beyond a shadow of a doubt, and I've seen (and saved copies of) countless screenshot comparisons myself, including fast-action scenes. Don't expect anybody in this thread to waste their breath trying to convince you; they have nothing to gain from that. The question nobody has a good answer to is why Amazon would reduce the quality of their VBR profiles when it does not appear to involve any reductions in bitrate or savings in bandwidth. It's also reasonable to question how a 1080p 14Mbps VBR stream can look worse than a 10Mbps CBR stream that is only 75% of the size. It's very counterintuitive, yet the differences are clear and consistent across films.

Given what I've mentioned, you can understand why some people would suspect that Amazon intentionally reduced the quality of their streams, and the "intention" is still unclear. If it wasn't intentional, then we'd have to assume that Amazon's encoders are ignorant and incompetent, and I don't think many people would make that assumption.

Edit: I mistakenly quoted EvilMonkey to make a point, so I took out the quote. This message was intended for users like 0x0x0x0x0.

 

[cartman0208]

... The question nobody has a good answer to is why Amazon would reduce the quality of their VBR profiles when it does not appear to involve any reductions in bitrate or savings in bandwidth. It's also reasonable to question how a 1080p 14Mbps VBR stream can look worse than a 10Mbps CBR stream that is only 75% of the size. It's very counterintuitive, yet the differences are clear and consistent across films....

I've got a good one (you will find, that I mentioned that earlier) If you have to recode tenthousands (millions?) of movies and TVshows quickly to reduce the bandwith of the suddenly increased streaming usage due to C19, you would do that with a lower quality profile to save time and CPU power

 

[jaqob]

I've got a good one (you will find, that I mentioned that earlier) If you have to recode tenthousands (millions?) of movies and TVshows quickly to reduce the bandwith of the suddenly increased streaming usage due to C19, you would do that with a lower quality profile to save time and CPU power

The problem with that explanation is that the current VBR streams don't seem to have lower bitrates/file sizes than the older ones from years ago (though I'm not 100% certain), so there wouldn't be any savings in bandwidth. As far as I know, the current ones tend to be max 15Mbps for 1080p. Has there ever been a time when the bitrates were higher than that?

 

[0x0x0x0x0]

It is a pretty well-documented fact in the p2p community that Amazon switched from a good VBR profile to worse profiles in the last few years, making CBR the highest quality stream. The files exist in the p2p community to prove this beyond a shadow of a doubt, and I've seen (and saved copies of) countless screenshot comparisons myself, including fast-action scenes. Don't expect anybody in this thread to waste their breath trying to convince you; they have nothing to gain from that. The question nobody has a good answer to is why Amazon would reduce the quality of their VBR profiles when it does not appear to involve any reductions in bitrate or savings in bandwidth. It's also reasonable to question how a 1080p 14Mbps VBR stream can look worse than a 10Mbps CBR stream that is only 75% of the size. It's very counterintuitive, yet the differences are clear and consistent across films.

Given what I've mentioned, you can understand why some people would suspect that Amazon intentionally reduced the quality of their streams, and the "intention" is still unclear. If it wasn't intentional, then we'd have to assume that Amazon's encoders are ignorant and incompetent, and I don't think many people would make that assumption.

Edit: I mistakenly quoted EvilMonkey to make a point, so I took out the quote. This message was intended for users like 0x0x0x0x0.

The problem with that explanation is that the current VBR streams don't seem to have lower bitrates/file sizes than the older ones from years ago (though I'm not 100% certain), so there wouldn't be any savings in bandwidth. As far as I know, the current ones tend to be max 15Mbps for 1080p. Has there ever been a time when the bitrates were higher than that?

And for the users "like you"---there is literally no way that a CBR would be anywhere near the quality of the VBR stream when the latter spends upto twice (or more) the bits of the former when scene complexity requires. The users "like you" can never quantify what "better" is, and like I said before, all users "like you" do is compare stills and occasionally you get a hair that is more pronounced in a CBR frame over a VBR frame (note that somehow the CBR-philes never compare high complexity frames, and I know why!)---not how one watches a motion picture at all! The longest a frame stays on the screen is 1001/24000 of a second, or 41.7 milliseconds, and that's what users "like you" simply do not get...

In fact, you can't even be bothered read and understand what was being asked: I never asked anyone to waste their bandwidth to prove anything, I expressly asked for source so I can verify the claims, and the only source I was provided had massively disparate bandwidth demand requirement that was higher in the VBR stream.

Finally, you're patently confusing a bandwidth requirement is vs stream bitrate, the peak bandwidth requirement is "availability of this much bandwidth is necessary for complete stream delivery." When mixed with VBR, you can transfer more bits of the stream into a buffer (cf. picture decode buffer) to ensure contiguous recovery-free playback. As for "Has there ever been a time when the bitrates were higher than [15Mbps for 1080p]?" sure, here's a show from 1993:-

...
Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 4 frames
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Duration                                 : 48 min 13 s
Bit rate                                 : 13.4 Mb/s
Maximum bit rate                         : 29.7 Mb/s
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 23.976 (24000/1001) FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.269
Stream size                              : 4.50 GiB (98%)
...

Good luck encoding that at even 20Mbps CBR and retain the details where VBR requires 29.7Mbps!

 

[zanetti]

Good luck encoding that at even 20Mbps CBR and retain the details where VBR requires 29.7Mbps!

wouldnt the solution be: raise the minimum bitrate of a VBR, rather than encoding a CBR with the max bitrate of a VBR? i guess that may defeat the purpose of VBR, lol.

 

[0x0x0x0x0]

wouldnt the solution be: raise the minimum bitrate of a VBR, rather than encoding a CBR with the max bitrate of a VBR? i guess that may defeat the purpose of VBR, lol.

VBR is not coded for bitrate but for scene complexity (in the old, like really old, days VBR was synonymous with ABR, but it's not been the case for a very long-long-long time)

 

[EvilMonkey]

Here we go again. Nobody is saying that CBR is better except for in Amazon's case. You really are wasting your time and not proving anything by comparing CBR and VBR if it does not contain both versions from Amazon.

 

[jaqob]

And for the users "like you"---there is literally no way that a CBR would be anywhere near the quality of the VBR stream when the latter spends upto twice (or more) the bits of the former when scene complexity requires. The users "like you" can never quantify what "better" is, and like I said before, all users "like you" do is compare stills and occasionally you get a hair that is more pronounced in a CBR frame over a VBR frame (note that somehow the CBR-philes never compare high complexity frames, and I know why!)---not how one watches a motion picture at all! The longest a frame stays on the screen is 1001/24000 of a second, or 41.7 milliseconds, and that's what users "like you" simply do not get...

In fact, you can't even be bothered read and understand what was being asked: I never asked anyone to waste their bandwidth to prove anything, I expressly asked for source so I can verify the claims, and the only source I was provided had massively disparate bandwidth demand requirement that was higher in the VBR stream.

@0x0x0x0x0, I don't doubt that you have a firm understanding of the differences between CBR and VBR in theory and practice. However, you must understand that when it comes to the differences between Amazon's VBR and CBR encodes you are clueless. If you weren't clueless and you had done your research, examining the screenshot comparisons for as many films as other users in this thread have, then you wouldn't continue trying to lecturing people about the general benefits of VBR over CBR, which are irrelevant to the discussion. And if you had paid attention to what I said, I mentioned "including fast action scenes", not just stills. It's not just the still that look better. Did you honestly think that the "CBR-philes" looked at a few stills and called it a day?

The example you gave has an average bitrate of 13Mbps, which is below the average bitrate I mentioned (15Mbps), plus I was talking about Amazon's encodes from 2-3 years ago, asking if they had higher bitrates back then, which I doubt. Most people don't have access to those.

"Patently confusing a bandwidth requirement is vs stream bitrate" - Again, I appreciate your willingness to lecture me, but that doesn't address what I was asking, which is why Amazon would switch VBR profiles if today's encodes appear to be just as sizeable as those from years ago.

 

[EvilMonkey]

I finally came across some videos that anystream grabbed the CBR

General

Complete name : Pandas Born to be Wild.mp4

Format : MPEG-4

Format profile : Base Media / Version 2

Codec ID : mp42 (mp42/iso6)

File size : 3.58 GiB

Duration : 53 min 6 s

Overall bit rate mode : Constant

Overall bit rate : 9 650 kb/s

Movie name : Pandas: Born to be Wild

Encoded date : UTC 2021-07-12 17:42:00

Tagged date : UTC 2021-07-12 17:42:00

Comment : Unlock the mysteries of wild pandas whose counterparts in captivity are known for their gentle image. Journey through the Qinling Mountains with filmmakers, scientists and rangers to witness pandas’ courtship and aggression behaviors.


Video

ID : 1

Format : AVC

Format/Info : Advanced Video Codec

Format profile : High@L4

Format settings : CABAC / 4 Ref Frames

Format settings, CABAC : Yes

Format settings, Reference frames : 4 frames

Codec ID : avc1

Codec ID/Info : Advanced Video Coding

Duration : 53 min 6 s

Bit rate mode : Constant

Bit rate : 9 455 kb/s

Nominal bit rate : 10 000 kb/s

Maximum bit rate : 20.0 Mb/s

Width : 1 920 pixels

Height : 1 080 pixels

Display aspect ratio : 16:9

Frame rate mode : Constant

Frame rate : 29.970 (30000/1001) FPS

Color space : YUV

Chroma subsampling : 4:2:0

Bit depth : 8 bits

Scan type : Progressive

Bits/(Pixel*Frame) : 0.152

Stream size : 3.51 GiB (98%)

Encoded date : UTC 2021-07-12 17:42:00

Tagged date : UTC 2021-07-12 17:42:00

Color range : Limited

Color primaries : BT.709

Transfer characteristics : BT.709

Matrix coefficients : BT.709

Codec configuration box : avcC

 

[0x0x0x0x0]

Here we go again. Nobody is saying that CBR is better except for in Amazon's case. You really are wasting your time and not proving anything by comparing CBR and VBR if it does not contain both versions from Amazon.

Don't switch the argument, you assert Amazon CBR is better than VBR, you only argue by reference to some third party sites with no verifiable provenance of the streams. Am I the only one who can see the scientific/engineering ignorance of that argument?

<snip>

The example you gave has an average bitrate of 13Mbps, which is below the average bitrate I mentioned (15Mbps), plus I was talking about Amazon's encodes from 2-3 years ago, asking if they had higher bitrates back then, which I doubt. Most people don't have access to those.

The 1993 show I "mediainfo-ed" was on Prime for the past three years, at least; I very much down Amazon have nothing better to do than to keep re-encoding old titles just for kicks.

"Patently confusing a bandwidth requirement is vs stream bitrate" - Again, I appreciate your willingness to lecture me, but that doesn't address what I was asking, which is why Amazon would switch VBR profiles if today's encodes appear to be just as sizeable as those from years ago.

Because VBR gives a better quality film (again, we're not encoding a JPEG at a constant QP here!). One thing you might have stumbled upon, as I'm sure already been mentioned is that, depending on your region, Prime might have removed top VBR entries for the streaming ladder due to a spike in demand during the house arrests. You keep repeating VBR looks worse than CBR, but literally no examples given; at least give me a title name so I can see "the terrible quality" of the VBR that AS would get. All your arguments are literally opaque "trust me"-kind. C'mon!

I finally came across some videos that anystream grabbed the CBR

<snip>

Are you sure it's actually CBR:-

...
Bit rate : 9 455 kb/s
Nominal bit rate : 10 000 kb/s
Maximum bit rate : 20.0 Mb/s
...

what's the bitrate histogram like?

 

[EvilMonkey]

If I just said you win would you finally shut up and stop trolling this thread?

 

[0x0x0x0x0]

If I just said you win would you finally shut up and stop trolling this thread?

I'm certainly not trolling the thread, but you are not providing any tangible evidence to support the thing you say... That's the problem, I am happy to be proved wrong, how else would I learn, the problem is that you are literally not providing any quality evidence to support your assertion. If I made a claim and said look at some dubious piece of evidence at some site X of unknown provenance, if you were a rational human being, you'd laugh your head off!

 

[EvilMonkey]

Bye!


Now I got to figure out how to block this thread.

Edit: found it, unwatched!

 

[Jimc115]

And for the users "like you"---there is literally no way that a CBR would be anywhere near the quality of the VBR stream when the latter spends upto twice (or more) the bits of the former when scene complexity requires. The users "like you" can never quantify what "better" is, and like I said before, all users "like you" do is compare stills and occasionally you get a hair that is more pronounced in a CBR frame over a VBR frame (note that somehow the CBR-philes never compare high complexity frames, and I know why!)---not how one watches a motion picture at all! The longest a frame stays on the screen is 1001/24000 of a second, or 41.7 milliseconds, and that's what users "like you" simply do not get...

In fact, you can't even be bothered read and understand what was being asked: I never asked anyone to waste their bandwidth to prove anything, I expressly asked for source so I can verify the claims, and the only source I was provided had massively disparate bandwidth demand requirement that was higher in the VBR stream.

Finally, you're patently confusing a bandwidth requirement is vs stream bitrate, the peak bandwidth requirement is "availability of this much bandwidth is necessary for complete stream delivery." When mixed with VBR, you can transfer more bits of the stream into a buffer (cf. picture decode buffer) to ensure contiguous recovery-free playback. As for "Has there ever been a time when the bitrates were higher than [15Mbps for 1080p]?" sure, here's a show from 1993:-

...
Video
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L4
Format settings                          : CABAC / 4 Ref Frames
Format settings, CABAC                   : Yes
Format settings, Reference frames        : 4 frames
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Duration                                 : 48 min 13 s
Bit rate                                 : 13.4 Mb/s
Maximum bit rate                         : 29.7 Mb/s
Width                                    : 1 920 pixels
Height                                   : 1 080 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 23.976 (24000/1001) FPS
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.269
Stream size                              : 4.50 GiB (98%)
...

Good luck encoding that at even 20Mbps CBR and retain the details where VBR requires 29.7Mbps!

Your logic is flawed at best.

The Average bitrate of what you displayed is 13.4 Mb/s.
Maximum Bitrate is what the Encoder was set too. It does NOT reflect an actual bitrate the video achieved. It is the theoretical
maximum the encoding program will allow to happen. It in NO WAY means that that bitrate was ever achieved.

Oh...and 1001/24000 is just a different mathematical way of expressing the FPS. 1001/24000 is equal to 23.976 FPS. It is the speed at which the film was recorded. It has NOTHING to do with VBR or CBR.

VBR streams normally carry a larger percentage of Audio, a reason why many uploaders will mux the HIGHER QUALITY CBR VIDEO stream with the HIGHER QUALITY VBR AUDIO stream.

 

[0x0x0x0x0]

Your logic is flawed at best.

The Average bitrate of what you displayed is 13.4 Mb/s.
Maximum Bitrate is what the Encoder was set too. It does NOT reflect an actual bitrate the video achieved. It is the theoretical
maximum the encoding program will allow to happen. It in NO WAY means that that bitrate was ever achieved.

Oh...and 1001/24000 is just a different mathematical way of expressing the FPS. 1001/24000 is equal to 23.976 FPS. It is the speed at which the film was recorded. It has NOTHING to do with VBR or CBR.

VBR streams normally carry a larger percentage of Audio, a reason why many uploaders will mux the HIGHER QUALITY CBR VIDEO stream with the HIGHER QUALITY VBR AUDIO stream.

Wow... let's deal with the easy parts first: virtually all streaming providers separate audio and video streams, so your last para is just utterly irrelevant.

Now the penultimate paragraph is a bit more difficult... Where did I say that FPS has anything to do with VBR or CBR??? Seriously, where??? Reading comprehension is really going downhill... If you're going to be arguing a point, can you at least bother to read what is written and comprehend it? As for 1001/24000 itself, here comes some maths and physics... As you recall, period of a wave is 1/frequency (basic school physics). So if the film frequency is 24000/1001 FPS, the period of each frame (which is the point I was making, if you actually bothered to read!) is 1/(24000/1001) seconds, or if you recall your elementary maths 1/(n/d) is equivalent to simply d/n, so you get your 1001/24000 seconds. What gave you the impression that it related to VBR/CBR is simply beyond me.

Now the difficult part... Go look at mediainfo's source code; the peak bitrate is computed, here's a computed output from ffmpeg using ffmpeg_bitrate_stats for the same file:-

   "stream_type": "video",
    "avg_fps": 23.976,
    "num_frames": 69385,
    "avg_bitrate": 13359.827,
    "avg_bitrate_over_chunks": NaN,
    "max_bitrate": 31134.355,
    "min_bitrate": 9.811,
    "max_bitrate_factor": 2.33,

You want to find out how ffmpeg_bitrate_stats computes max_bitrate, it's on github!


Anything else?

 

[zanetti]

VBR is not coded for bitrate but for scene complexity (in the old, like really old, days VBR was synonymous with ABR, but it's not been the case for a very long-long-long time)

sure, but if minimum bitrate is high enough to accomodate even the most complex of scenes in a movie... ?

 

[0x0x0x0x0]

sure, but if minimum bitrate is high enough to accomodate even the most complex of scenes in a movie... ?

That's an absolutely fine decision, so long as you're prepared to spend 31134.355 Kbps on scenes that need merely 9.811 Kbps (for example, see above example), video coding is always an optimisation problem [1] defined by your parameters (time-to-encode, time-to-decode, quality, and bit rate; pick whatever balance you want between the four)

1. https://en.wikipedia.org/wiki/Optimization_problem