Thanks for AV1 encode. ❤️

No problem, if anyone has requests for AV1 encodes, suggest them. However keep in mind that the series should have finished airing.

Can you do beatless?

Yeah, I don't see why not.

感谢分享！

do you plan on uploading the violet evergarden movie?

Hard question, I don't currently deal with Movies atm, but maybe other HAV1T member could do that for you.

Need violet evergarden movie https://nyaa.si/view/1230203 Also what are the visual difference between av1 compared to other encodes?

AV1 has insane compression efficiency, but is very slow to encode. Visually it really likes to denoise, so to avoid banding I must use 10bit. Artifacts are hard to notice, it hides them very well, however it is untuned for dark scenes, so it tends to drop bitrate too much where it shouldn't. So for now AV1 is great for high quality lossy compression, but if you are looking for transparency, h264 can't be beaten in that regard as of yet, however you will need to sacrifice a lot of disk space and internet bandwidth.

Thanks for the information. It sounds good and all but its no use to me if I can't play the videos on my phone. The frames lag every few seconds. Using samsung phone with latest vlc player.

Yeah, AV1 has a lot of kinks to iron out. But since the development on the decoder has accelerated, I don't see the point of encoding in 8bit, because I will have to redo it anyways.

Doumo arigatou Montec-san! <3

thanks for the release, please try to include dub too, I mux your files with dub for myself but ep13 doesn't match with any dub release

Indeed, episode 13 has subtitle desync issue, probably a weird release difference between localized blurays. We are trying to improve quality control, and I'll make sure to solve them. I also must add how painful it was encoding Violet Evergarden. It has amazing animation, but there is one notable problem. It really likes post filters on top of that animation, so there are huge amounts or gradient colors and dark scenes. Gradient colors isn't much of an issue, but the dark scenes... This is (if I remember correctly) the 6th time I encoded Violet Evergarden. I wasted close to a month of processing power on i7 6700 to get it to proper quality without making it overly bloated. It was not possible to make it 8bit and look good because banding was insane, and artificial grain did not help at all. Also due to lack of documentation all of the testing had to be done by HAV1T team. There is an insane amount of work being done behind the scene. We even developed our own tools to improve encodes because developers didn't listen to us. What can I say is this: this is not even close to AV1 limit, if I used the best possible settings, I think I could get 720p video (anime content, 24 minutes) without any visible artifacting to 50MB.

The issue is more like FPS issue, if you encode it from another source it should be fine,50mb 720P for AV1 is kinda wasted, when you encode 2 frame/sec I watch it and keep it, I do not watch it and then delete it

This is wayyy too overkill, lol. not event my 8th gen i5 laptop can play it normally when a heavy movement is going on oh well, at least updating mpv helps up a bit

@Nexxkinn my 4th gen dual core i7 laptop can play it with ease,try to play it with vlc player, I am using arch I build mpv from aur with all hardware accelerator and stuff, it still drop frames

Missing on the description: Sound is opus@96kbit. While opus is amazing, the bitrate is too low. It doesn't need to be flac, but it wouldn't hurt to dedicate 160kbit/s and make it transparent to most listeners. I'm getting a different release just for the sound.

We did blind testing with original .wav and opus files. Compressed with opus, then back to .wav so people couldn't tell which file is which. Sample size was 6 people in total, 5 out of 6 preferred opus. Also other studies were done with more people, and it was 50/50 when it came to telling which one was the compressed file. Opus at 96kbps is the optimal bitrate for very good quality and good file sizes. However a placebo effect is a real thing and if you know before hand which file has which bitrate, your brain will become biased. @kanone However if you do hear a noticeable difference at a certain scene I would really appreciate the time stamp.

@Montec >However a placebo effect is a real thing and if you know before hand which file has which bitrate, your brain will become biased. I'm well aware of double blind testing. And the most recent large scale double blind testing experiment done at hydrogenaud.io shows OPUS is the best of the codecs tested, for all codecs tested at 96kbit. OPUS is awesome. There's seldom justification for using any other lossy codec. However, this testing does not say anything about 96kbit being transparent. I recommend you experiment with the recordings on the OPUS demo pages. https://people.xiph.org/~xiphmont/demo/ You'll be able to experiment and see (or rather, hear) by yourself that 96kbit isn't the same as lossless. You'll find many specific points at the samples where the difference will be reliably apparent to your brain, through your ears and equipment. An added 32kbit might not mean much to video; this is fact insignificant when compared to the video bitrate. But it can make a huge difference in audio. Why not play it safe with a bitrate that has a chance to be transparent to most people? Xiph themselves (they're behind opus) do recommend 128kbps for transparency: https://wiki.xiph.org/index.php?title=Opus_Recommended_Settings&mobileaction=toggle_view_desktop

Because I cannot hear any difference, neither other members of our group. Give us a timestamp so we could find the loss of quality you're complaining about. Also adding additional 32kbps is not insignificant when we are talking sub 200MB files, in some cases sub 100MB. For every episode this is an additional 5.76MB (assuming it is 24 minutes long). So let's say we have 24 episodes with that additional 32kbps, that would equal to 137.76 MB to the overall file size, which could be used for additional OVA or an episode. Adding bitrate just because it seems right doesn't mean it is right.

@Montec, You seem to be fixated on personally hearing differences. I gave you a link to the Xiph demo pages, where you can experiment with listening to samples encoded at different bitrates until you're sick of it. But my take is that the subjective experience of a single person is useless, which is why large scale double blind tests are conducted. This research has been done methodically by those who take it very seriously, and there's no point arguing over our circumstantial evidence based on just our ears and equipment. It boils down to: Audio transparency is, in terms of bitrate cost, extremely cheap when compared to video transparency. Do you care about audio transparency? If you do, you use 128kbit as recommended by Xiph. If you don't, you do continue to sacrifice that additional 32kbps, and perhaps fit that extra OVA within the same overall download size, at the cost of transparency. Or, if you really want to still fit in that extra OVA without losing audio transparency, you could take 32kbit from the video, and give it to audio. Of course, this is hypothetical. I'm not saying to go and do it, considering the current cost of AV1 encoding. Honestly speaking, I'm not even saying remux the audio, flush the torrent down the toilet and make a v2. It is just a thought exercise. Please consider 128kbit OPUS for future encodes. Thank you very much for the AV1 encode.

First I thank you for your efforts. I found that the videos are not _tiled_, well, I assume you are using `aom` or `rav1e` encoder(whether ffmpeg built-in or aomenc). Tiling is very important for AV1 since it would boost the decode speed a lot with just a little increase in bit-rate. It also enables the encoder's multi-threading capability. 4x2 tiles is quite a good choice and YouTube is also using it. How do you think? :)

We could still use an 8bit version because I know I can not play 10 at this resolution. Even if you could, you may not want your computer fans making so much noise.