The cutoff option sets the bandwidth of the encoder – bandwidth meaning the highest frequency that will be encoded.
While typically you wouldn't restrict the bandwidth when encoding audio for home theater, music listening, etc., there are other use cases such as telephony or video conferencing, where it is not necessary to transmit the whole frequency spectrum.
For example, the primary frequencies of speech are situated well below 8 kHz, so you can cut off the higher frequencies and save required transmission bandwidth. This was also done in the early days of telephony to be able to share a line among multiple users.
Various encoders support this option, such as:
Especially AAC and Opus are interesting because they were created with low bit rate or narrow band transmissions in mind.
2
I think for AAC they're advocating setting it to 15,000 – which should still be enough for most music. The encoder may save bandwidth and you might lower the risk of encoding hissy frequencies above that, which could cause artefacts. There's also some psycho-acoustic trickery going on, but I'm not sure which encoder uses that.
– slhck – 2013-04-01T09:37:22.8632
@StevenPenny Also see Quality Assessment of FFmpeg AAC for a graph showing how various
– llogan – 2013-04-01T18:09:11.743-cutoff
values may affect perceived quality for the native FFmpeg AAC encoder (-codec:a aac -strict experimental
).@LordNeckbeard It's quite interesting how the MOS drops for higher cutoffs – not what you'd really expect. – slhck – 2013-04-01T18:58:04.937
1@slhck it makes sense to me. As you explained a higher cutoff increases the likelihood of artifacts from high freq tones. – Steven Penny – 2013-04-02T03:28:46.120