[opus] Is OPUS_AUTO the default for an encoder's bitrate?

[Kevin O'Connor]

I had expected that the default bitrate for the encoder would be the same as setting it to OPUS_AUTO, but I'm getting difference results:

>opusenc --comp 4 sample.wav sample.opus
Encoding using libopus 1.3-rc2 (audio)
-----------------------------------------------------
   Input: 8 kHz, 1 channel
  Output: 1 channel (1 uncoupled)
          20ms packets, 25 kbit/s VBR
 Preskip: 312

Encoding complete
-----------------------------------------------------
       Encoded: 2 minutes and 34.94 seconds
       Runtime: 1 second
                (154.9x realtime)
         Wrote: 588189 bytes, 7747 packets, 157 pages
       Bitrate: 29.7104 kbit/s (without overhead)  Instant rates: 1.2 to 50.4 kbit/s
                (3 to 126 bytes per packet)
      Overhead: 2.17% (container+metadata)

If I explicitly set the bitrate to OPUS_AUTO, it produces a stream with twice the bitrate of the encoder's default from above.

>opusenc --comp 4 --set-ctl-int 4002=-1000 sample.wav sample.opus
Encoding using libopus 1.3-rc2 (audio)
-----------------------------------------------------
   Input: 8 kHz, 1 channel
  Output: 1 channel (1 uncoupled)
          20ms packets, 25 kbit/s VBR
 Preskip: 312

Encoding complete
-----------------------------------------------------
       Encoded: 2 minutes and 34.94 seconds
       Runtime: 1 second
                (154.9x realtime)
         Wrote: 1245829 bytes, 7747 packets, 157 pages
       Bitrate: 63.6644 kbit/s (without overhead)  Instant rates: 1.2 to 120.4 kbit/s
                (3 to 301 bytes per packet)
      Overhead: 1.03% (container+metadata)

As this is example 8kHz single channel audio, the encoder's default bitrate seems more appropriate than the OPUS_AUTO bitrate. This presents a bit of a problem because there are three possibilities for the bit rate, an explicitly set value, a value set by OPUS_AUTO, and some mystery value from the default encoder initialization. If the bitrate setting is changed, I don't see any way to get back to the mystery value without re-initializing the encoder.  Can anyone comment on why it works this way?

Kevin O'Connor