ffmpeg -i file.wav -ab 320k newfile.opus
I don't know how to decide the value to set for the -ab option. But my overall goal is to convert wav file to opus file with high quality, and delete the wav files to save disk space. How to determine what bit rate is high enough and still "relevant"?
You have some recommendations here :
https://wiki.xiph.org/Opus_Recommended_Settings
For 2 channels, Opus at 128 KB/s (VBR) is pretty much transparent.
High enough may be subjective…
For most samples, you'll get very close to transparency for the typical setup with as low as 96 kbps. This is the bitrate where Opus really shines and is often a good trade-off between quality and size, while heavily leaning to quality. Above that, you get very rapid diminishing returns.
But let me tell you what "transparent" means in reality. When people use double-blind testing (like the ABX method) to know if a lossy codec is indistinguishable from lossy they aren't just listening to the music and noticing that it sounds worse. They are putting on expensive high-quality headphones and listening to the same 1 second clip of cymbals (or another part of the music that is infamously hard for lossy codecs to reproduce at low bitrates) over and over and over at high volumes and over until they notice tiny, tiny differences between the lossy and lossless copies. Once they can no longer tell a difference, it is considered transparent.
Btw the answer by Frédéric recommends 128 KB/s. I assume that's a typo and he means 128 kbps (128 KB/s would be 1024 kbps). This is also a good bitrate. Going too much above this bitrate, other high-quality audio codecs such as AAC and Vorbis start to be competitive and you largely lose the advantage of Opus. For stereo music I would never go above 160 kbps. 144 kbps would be my practical limit.
If you want to save space without any loss in quality (even imperceptible loss), you can cut your WAV collection's size in half (40-60% savings typical) by compressing it with a lossless codec like FLAC. I always use FLAC for long-term audio collections because sometimes I have to convert the audio into a different format to make it compatible with a device that doesn't support Opus (and converting lossy to lossy is always bad).
