**Description**
The video is from VCB-Studio. They applied debanding and slight AA, plus adaptive denoising and texture sharpening. Both the remastered 5.1 and original 2.0 audio are from AK-Submarines, but I encoded them to Opus. I set 512kbps for the 5.1 and 192kbps for the 2.0. This cut the audio files to around 1/8 of the original 24bit audio, saving 9GB total. The main subs are the improved /ak/ Submarines subs from 2019. Also included are Hiryuu, Commie (Übermensch), and Commie (No fun allowed) subs. I directly copied the chapters from AK-Submarines so these should work with their optional ordered chapter supercut files. This is a smaller audio version of my [FLAC release](https://nyaa.si/view/1335652).
**Info**
Video: 1920x1080p x265 10-bit
Audio 1: Opus 5.1 512kbps Remaster
Audio 2: Opus 2.0 192kbps Original
Sub 1: /ak/ Submarines (improved, great signs)
Sub 2: Huryuu
Sub 3: Commie (Übermensch)
Sub 4: Commie (No fun allowed)
Chapters: yes
**Tech Note**
When encoding from 24bit audio to lossy codecs you shouldn't dither since the lossy codec's decoder already applies dither. I enabled the (non-default) improved Opus souround encoding in ffmpeg with -mapping_family 1. To begin, libopus 1.1+ allocates bitrate in a more desirable way instead of spreading it evenly. For 512kbps, libopus starts by allocating 170kbps for L+R, 170kbps for Lb+Rb (Ls+Rs must get remapped), 85kbps for C, and 77kbps for LFE. It then uses surround masking which takes advantage of cross-channel masking between free-field speakers. This of course is done in the context of retaining stereo downmixing. It also analyzes each channel's relative audible contribution within the sound field per critical band, and adjusts bitrate accordingly. After all the optimizations the average bitrate turns out to be around 400kbps. I chose 512kbps and 192kbps because they're a great balance between file size and quality. Those values are smaller than what's being used for AAC yet Opus has superior quality. The tradeoff for anything smaller isn't worth saving a handful of megabytes.
TL;DR Proper Opus and FLAC sound the same.
>After all the optimizations the average bitrate turns out to be around 400kbps
Wonder if that's why when I switched from FFmpeg based encoding to the official opus encoder the average bit rates reported by mediainfo went down? I could be retarded though.
"~494 kb/s" on my guilty crown encode vs "~413 kb/s" on my more recent release
---
>Those values are smaller than what’s being used for AAC yet Opus has superior quality.
Does this apply to VBR Q 127 QAAC as well?
> Wonder if that’s why
Yeah that's why. opusenc enables the new surround encoding by default. ffmpeg turns it off by default.
> Does this apply to VBR Q 127 QAAC as well?
Yes. q127 QAAC targets 320kbps for stereo and 768kbps for 5.1. Opus is usually transparent at 160kbps for stereo but some harder samples need 192kbps, which is why I stick with 192kbps. For 5.1, 448kbps is usually transparent but again some harder samples might need a bit higher so 512kbps target is perfect. Opus 1.3 even uses a recurrent neural network to dynamically switch back and forth between using speech optimized and music optimized encoding. With look-ahead the decision logic is even able to have the encoder switch modes during a silence period before the transition between speech and music. Opus is the highest quality widely available audio codec. It's also open source unlike QAAC (which is just a wrapper around Quicktime AAC) if you care about that.
>Yeah that’s why. opusenc enables the new surround encoding by default. ffmpeg turns it off by default.
That's good to know, I was worried I was getting lower quality files because they came out 10MiB smaller than what I was getting from FFmpeg based encoders. I figured it was probably fine though considering it's the official encoder.
Comments - 4
Simplistic
TechNinja (uploader)
Simplistic
DoktorWeiss