I encoded a stereo track using libopus 1.2 alpha in foobar2000, and then once more with the same track except I downmixed stereo to mono before encoding. I used 48 kbps for my target bitrate. Both the stereo and mono encoded files came out around 48 kbps. I tried again at 128 kbps (just to test at a significantly higher bitrate) and the results were still at or close to the target bitrate for both files.
I was not expecting this since LAME, qaac, and Vorbis all tend to cut the bitrate accordingly if it detects very little or no stereo separation. Does Opus not operate in a "joint stereo" mode?
Of course it does. It just targets the bitrate you specify, regardless of how many channels the signal has, or how simple or complex the inter-channel correlation may be. Easier to encode signals may end up higher quality than those which are harder to encode, while yielding the same bitrate.
It's not that big of a deal now that I know, but I'll just have to remember to knock the bitrate down to about 32 kbps for mono speech/podcasts and leave the stereo stuff at 48 kbps.
For batch encoding, it would be nice if there was a consistent quality setting like there is in LAME, Vorbis, etc.
Are there plans for a "True VBR" so to speak? The current VBR behavior is more like what I would expect from a "Constrained VBR". I've encoded some more samples around 48 kbps and, while Opus is excellent at these low bitrates, it's cumbersome to have to examine each file and decide whether it's complex enough to warrant bumping it up to 64 kbps. Similarly, I have to check which ones are mono and then knock them down to 32 kbps. A quality setting is definitely one advantage HE-AAC currently has over Opus.