[Icecast] Transcode lossy to further reduced lossy to stream over Icecast

D.T. ohnonot-github at posteo.de
Sun Apr 16 17:08:24 UTC 2023


PS: I used 64kbps as bitrate, so: br=64000
This creates comparable filesizes for both codecs, 25s samples come out
~200K.

On Sun, 2023-04-16 at 17:05 +0000, D.T. wrote:
> I created some test samples and transcoded to FDK AAC and libopus at
> fairly low bitrates - I cannot recreate what bothered me about Opus &
> noisy music previously.
> It also seems I cannot tease ffmpeg into encoding FDK's AAC with VBR.
> As it stands, Opus clearly wins in this scenario.*
> 
> Q:
> Is it possible to stream in variable bitrate?
> 
> 
> *
> ffmpeg -i "$track" -vn -ac 2 -c:a libfdk_aac -profile:a aac_he_v2 -b:a $br -vbr 5 "${track}.HEv2_$br.aac"
> ffmpeg -i "$track" -vn -ac 2 -c:a libopus -b:a $br -vbr on "${track}.vbr$br.ogg"
> 
> 
> On Sat, 2023-04-15 at 16:17 -0500, Brad Isbell wrote:
> > Opus or AAC will give you comparable results at reasonable bitrates
> > (~128k).
> > Though, I would suggest finding a way to get more storage.  You
> > could upload to Backblaze B2 or AWS S3 for pennies, if your current
> > host won't let you upgrade.
> > 
> > On Sat, Apr 15, 2023 at 3:36 PM D.T. <ohnonot-github at posteo.de>
> > wrote:
> > > Situation: 
> > > * remote virtual server with very little storage (estimate: I can
> > > spare about 40G for music)
> > > * local music collection of ~80G in all sorts of formats - lossy
> > > in
> > > varying quality, some lossless too
> > > 
> > > Vision:
> > > * stream my whole music collection randomized so I can listen to
> > > it
> > > anywhere
> > > 
> > > Plan/Idea:
> > > * Locally transcode everything to one format that results in
> > > files
> > > that are 
> > > - small enough to fit on my server, altogether
> > > - have a reduced bitrate for streaming
> > > - can be streamed as-is without further transcoding
> > > * Upload
> > > * Set icecast up to do just that (this I know how to do)
> > > 
> > > So I'm asking advice for the transcoding. What's likely to give
> > > the best results with already lossy sources, and at small
> > > bitrates?
> > > 
> > > According to these documents:
> > > https://trac.ffmpeg.org/wiki/Encode/HighQualityAudio
> > > https://wiki.hydrogenaud.io/index.php?title=Transcoding#Lossy-to-lossy_transcoding
> > > it basically comes down to Fraunhofer Institute's FDK AAC, but
> > > the articles are dated.
> > > Opus is supposed to be good but I always have the impression it
> > > doesn't deal well with loud/grungy/fuzzy/guitarry music, esp. at
> > > low bitrates.
> > > 
> > > What do you say?
> > > 
> > > _______________________________________________
> > > Icecast mailing list
> > > Icecast at xiph.org
> > > http://lists.xiph.org/mailman/listinfo/icecast
> > _______________________________________________
> > Icecast mailing list
> > Icecast at xiph.org
> > http://lists.xiph.org/mailman/listinfo/icecast
> 
> _______________________________________________
> Icecast mailing list
> Icecast at xiph.org
> http://lists.xiph.org/mailman/listinfo/icecast

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xiph.org/pipermail/icecast/attachments/20230416/d727b4c1/attachment.htm>


More information about the Icecast mailing list