thanks Dude ._.

still slower than Dragster

Thank you for this release!

Why is episode 1 only half the size of the rest?

because it's netflix, and netflix sucks

^That doesn't make any sense

It makes sense if you check netflix bitrate for Violet Evergarden or every anime on netflix.

Netflix is not that good at encoding.

He's saying that doesn't really make much sense as an explanation as to why episodes 2 - 9 are much larger than 1, 10, 11 and 12

^Exactly

Because netflix is bad at encoding.

Netflix is great at encoding, but their goals in encoding are different. Streaming is more important to them, so size varies based upon the amount of quality they can get per bit rate. They have a fancy algorithm to determine bit rate per episode that tries to get a certain quality level for each episode, using only as much bitrate as necessary to get there. They also need better seeking, so they keyframe every two seconds, which also increases the size of their encodes.

>They have a fancy algorithm to determine bit rate per episode that tries to get a certain quality level for each episode, using only as much bitrate as necessary to get there. yeah, except the quality they target differs per episode so you have instances like this where it drops a shitload for no real reason. just admit they're bad and don't know how to encode.

Dogflix strikes again

The algorithm explanation actually make sense, thanks

When you use CRF in video encoding, you will get variable size depending on the content of the anime. If the show has little action and not so complex pictures, then the size will be smaller. If the anime has some much action and complex pictures with many shades of colors, the the size will be big.

Here's an explanation of Netflix's Per-Title Encode Optimization: https://medium.com/netflix-techblog/per-title-encode-optimization-7e99442b62a2 They're using VMAF now to determine the quality. The minimum quality they target is the same per title, so when the bitrate drops, that's because the algorithm is determining that they don't need a higher bitrate for the minimum quality. And in the case of anime, you often have a very complex opening sequence that requires a higher bitrate than the rest of the episode.

>When you use CRF in video encoding, you will get variable size depending on the content of the anime. If the show has little action and not so complex pictures, then the size will be smaller. If the anime has some much action and complex pictures with many shades of colors, the the size will be big. yes, but they don't use CRF, which would work perfectly for what they want to do. for instance, with the same CRF, repeated parts like the OP or ED would always look the same, yet with how netflix encodes that isn't the case. those parts can have a huge difference in quality between episodes. >The minimum quality they target is the same per title, so when the bitrate drops, that’s because the algorithm is determining that they don’t need a higher bitrate for the minimum quality. except their minimum level of quality is apparently far too low. and regardless, they basically just made a worse version of what CRF already does. >And in the case of anime, you often have a very complex opening sequence that requires a higher bitrate than the rest of the episode. it doesn't end up working that way, though. i guarantee if you compare the OP between episode 1 and 2, the difference will be massive.

"In particular, we utilize VMAF, the Netflix subjective video quality metric, as our optimization objective, since our goal is to generate streams at the best perceptual quality." "Subjective video quality." "Best perceptual quality." It's all just bullshit. Their metric is based on the Average Joe who thinks that resolution is all that matters and that buffering is the worst thing that could happen.

In summary, netflix sucks.

Netflix's such a meme.

arigato ^.^

> yes, but they don’t use CRF Actually, they do. Their first pass is a CRF encode, whose bitrate is then passed along as the bitrate for the second pass encode if it is below their maximum bitrate for that resolution. For more info, *See* De Cock, Jan, et al. "Complexity-based consistent-quality encoding in the cloud." IEEE International Conference on Image Processing (ICIP), 2016). I'm sure it has been tweaked since then. > except their minimum level of quality is apparently far too low ¯\\\_(ツ)\_/¯ It's higher than CR, lower than Amazon, and acceptable for most shows. > i guarantee if you compare the OP between episode 1 and 2, the difference will be massive. Just did. You were right, the difference was massive. There's no OP in episode 1. There's fog in later episodes, which blows up the bitrate.

¯\_(ツ)_/¯ It’s higher than CR, lower than Amazon, and acceptable for most shows. > https://i.imgur.com/3ICKTPU.png not sure that it's better than 480p CR. https://i.imgur.com/0yHPLUH.png isn't really better. And in every shows by Netflix or on netflix, it happen.

I wasn't talking about bitrate, I was talking about quality. Higher bitrate ≠ higher quality, and past certain thresholds, there are diminishing returns to visual quality with additional bitrate. Violet Evergarden doesn't need that much bitrate to look good. Netflix's encode isn't bitrate starved on it, and I'm guessing the baseline bitrate is determined by a CRF 16 encode from the source. It's not great or generous, but it is reasonable and most people wouldn't be able to tell the difference between the Netflix encode and the source. For A.I.C.O. - Incarnation, it looks like Netflix set the maximum allowed bit rate on 1080p to 7500 kbps, which it hit on 3 episodes. That's a reasonable hard limit for 1080p streaming, and as far as I could tell in my quick skim, there aren't any visual artifacts in the encode. In contrast, CR hasn't ever cared about quality, as the quality fiasco from last year showed. Their 720p video bitrate was 1776 kbps, their 1080p bitrate was 3072, and too bad if that isn't enough.

You mean this quality? It sure looks good. https://i.imgur.com/vXSVLf0.png

much worse is they are subtitles missing sometimes

>Actually, they do. Their first pass is a CRF encode, whose bitrate is then passed along as the bitrate for the second pass encode if it is below their maximum bitrate for that resolution. For more info, See De Cock, Jan, et al. “Complexity-based consistent-quality encoding in the cloud.” IEEE International Conference on Image Processing (ICIP), 2016). I’m sure it has been tweaked since then. then how do you explain bitrates that are round numbers like exactly 1000kbps consistently? i know they have a max of a certain value, but this is way lower than it. >¯\_(ツ)_/¯ It’s higher than CR, lower than Amazon, and acceptable for most shows. no, it's even worse than CR. it's probably the worst streaming site for anime. >Higher bitrate ≠ higher quality, and past certain thresholds, there are diminishing returns to visual quality with additional bitrate. maybe if netflix didn't use awful x264 settings. >Violet Evergarden doesn’t need that much bitrate to look good. https://diff.pics/jDdCXCYjf3Oh/1 >In contrast, CR hasn’t ever cared about quality, as the quality fiasco from last year showed. Their 720p video bitrate was 1776 kbps, their 1080p bitrate was 3072, and too bad if that isn’t enough. just because HS still rips CR's shitty encodes because they're smaller doesn't mean that's still true. virtually everyone who streams video on their site gets the new higher quality encodes that other groups rip. you should probably stop talking about things you don't understand.

>then how do you explain bitrates that are round numbers like exactly 1000kbps consistently? I can't, because I've never seen that. What I have seen is that each step up in their 1080p bitrate ladder is around 5 percent, rounded to the nearest ten in kbps. For this series, the bitrates are 3160, 3830, 4220, 5800, 6190, and 7500, but I've also seen 6810, 5360, 5110, 4870, 4640, 4020, and 3320. They have different max thresholds for lower resolutions. But then again, I'm not looking at Netflix encodes below 1080p. As for the pics, you can pull individual frames out of video encodes, but it doesn't tell you much about the quality of the *video*. And here you really are comparing apples and oranges, when you're comparing a 720p upscale to a 1080p encode, when the original was mastered in 1080p. As for CR, my comments stand. The higher quality encodes don't stick around. You should probably stop talking about things you don't understand. But then, if you did, you wouldn't be herkz.

>As for the pics, you can pull individual frames out of video encodes, but it doesn’t tell you much about the quality of the video. it does when every frame is worse. >And here you really are comparing apples and oranges, when you’re comparing a 720p upscale to a 1080p encode, when the original was mastered in 1080p yet the upscale is still better than shitflix 'native' 1080p? why are you shilling for netflix on a torrenting website?

So netflix encoders are using magic algorithms to know how much bitrate the video needs to look good, but when you open the video, an upscale from tv720 of a native 1080p anime looks better than their magic 1080p encode. I don't think you need to understand things if you just open your eyes and take a look at the video.

>I can’t, because I’ve never seen that. wow, you definitely seem qualified to discuss this topic. >And here you really are comparing apples and oranges, when you’re comparing a 720p upscale to a 1080p encode, when the original was mastered in 1080p. doesn't it say a lot that the upscaled 720p looks better? >As for CR, my comments stand. The higher quality encodes don’t stick around. they do. shows how much you know. just give up.

herkz already lost the battle, live to fight another day my dude

Thank you sir ! :)

When will come the Blueray?