[opus] Antw: Re: Q: Bandwidth vs. bitrate
Ulrich.Windl at rz.uni-regensburg.de
Mon Nov 4 07:10:00 UTC 2019
>>> Mathias Buhr <napcode at apparatus.de> schrieb am 01.11.2019 um 10:28 in Nachricht
<8c9ae155-c066-0a85-4a04-41a3d5cb14be at apparatus.de>:
> Hi Ulrich,
> Other than the bitrate, none of the discussed settings should have a
> significant impact on the file size as the encoder will always use the
> available bitrate to encode the input signal. The additional settings
> only configure where the encoder spends the bits and how much time it
> has to figure that out.
> From my experience and as others noted, 15-32 kbps should yield very
> good quality for wideband (8kHz acoustic bandwidth) speech. Less than
> 15kbps still yields good and intelligible speech but can also bring some
> noticeable artifacts (obviously depends on the input). The complexity
> setting basically configures the trade-off between runtime and encoding
> quality (whatever that is). As you probably don't care about runtime,
> you should set it to 10 as this should yield the best quality for a
> given bitrate (the default is 10). This should actually allow the
> encoder to perform a signal analysis (done for >= 8, if I remember
> correctly) and determine the "best" bit allocation.
> Now if the noise contained in your signal has some energy in the upper
> bands, the encoder might try to encode that as well (assuming 48kHz
> input). In order to avoid that, the codec control OPUS_SET_BANDWIDTH
> might have been helpful.
The above is very helpful (for me who doesn't know all the details inside the black box). Maybe some of these "insights" could be added to the manual page for opusenc. Re-reading the manual, I also think that "--comp" that is said to default to 10 now, had had a different default in the past, but I may be wrong.
Probably I thought "complexity" would affect the size of the resulting file...
> opus mailing list
> opus at xiph.org
More information about the opus