Depending on the type of client devices, you may see less often transcoding if you go down to 224 kbps for the audio bitrate.
The quality is still perfectly fine for stereo, if you keep the libfdk_aac parameter.
1 Like
Your right. This has been on my to do list to go back and revisit for a while. 256k has always been my default for stereo and its worked well. Stream size is small for audio so its never really been an issue for me.
I do however have one user (my brother) who pretty much always watches plex via a browser remotely. This results in transcoding and I’ve suspected for a while it might be the audio. I’ve been considering going to 224k or even 192k but need to read up a bit more on it.
If you do a lot of downmixing from surround to stereo, this can improve the output considerably: How can I normalize audio using ffmpeg? - Super User
If you don’t do something like the above, you are better off ripping and using the already downmixed stereo track (which unfortunately is often missing, even on disc-sourced media).
Thank you. I’ll head down that rabbit hole once I’m 100% confident and happy with the video parameters I’m adjusting. I’ll go even more crazy than I already am if I start to add audio adjustments into my mix ATM.
That will be interesting. Hope you post your findings then too 
I know it isn’t on topic, but why are you guys downscaling TV shows and converting audio into stereo? I’m just curious. Is it only a matter of disk space?
Mainly disc space, yeah.
And also the fact that I am consuming most series when already in bed, where surround is utterly useless and will annoy both me and the neighbors with the rather large dynamic range it has.
So I’m down mixing and normalizing. Saving a lot of storage and bandwidth while doing so, and increasing the likelihood of Direct Play even over a remote connection.
As for the picture: my Shield does a very decent job of upscaling it back to full HD. I am avoiding the rabbit hole of 4k as long as I can.
1 Like
An additional remark on downscaling:
I’m collecting quite a bit of Anime. I’ve come to the conclusion that in the overwhelming majority of cases the image sharpness and the amount of detail in the original bluray rips doesn’t justify anything higher than 720p.
In some cases I’ve even seen that the upscaled picture from my Shield (720p->1080p) looks better than the original rip (1080p).
This might be changing now with brand-new productions, but it is very much valid for all the existing stuff from before the 2020’s.
1 Like
For me movies are 1080p but for shows I’ve always scaled down to 480p with Stereo (now 720p after this thread). Space & compatibility across various devices are the main reasons. Really we don’t need 1080p quality for shows.
I’ve a fair bit of anime also both old (80s) and modern. Your findings are interesting and I’ll keep this in mind as I move forward.
On the normalizing 
I was looking at the link you provided above and also having a look at this thread Sharing some ffmpeg scripts I made to down convert audio in batch as well a 'night mode / DRC' and was going to start my testing using -filter:a "acompressor=ratio=4:makeup=2,loudnorm". Would this be a good starting point ? I’d like to not use ffmpeg-normalize unless I had to as I want to keep everything in the one ffmpeg command.
1 Like
acompressor is actually applying compression, not just normalization.
The former is affecting the dynamic range, the latter only raises the volume (as long as this doesn’t cause clipping).
Compression should only ever be used on “speech”-type of content. If you use it on music, you might as well tune into FM radio.
1 Like
Been busy with real world personal stuff lately but did manage to get some time over the weekend to test out a few additional things.
On your recommendation Otto I switched the audio over to -codec:a libfdk_aac -ac 2 -filter:a loudnorm -b:a 224k -ar 48000 and this is working well. Thanks for your comments and guidance on this.
I’ve also switched over to using scale=1280:-2:flags=lanczos+accurate_rnd+full_chroma_int on my scale filter. Adding full_chroma_int has really helped with colour scale, especially when dealing with 422 or 444 input. See Reddit - Dive into anything for some additional comments if your interested.
Anyhow, for reference this is now my new downscaling to 720p command:
ffmpeg -y -i file:"input" -loglevel error -stats -map_metadata -1 -codec:v libx264 -vf "scale=1280:-2:flags=lanczos+accurate_rnd+full_chroma_int,format=yuv420p" -crf 20 -profile:v high -level 31 -preset slow -metadata creation_time=now -codec:a libfdk_aac -ac 2 -filter:a loudnorm -b:a 224k -ar 48000 -sn -dn -f mp4 file:"output"
