Why did you dither the audio to 16 bits? Bit depth is only meaningful to Opus when it comes to processing bits underneath the noise floor, and it's generally considered best practice to leave any source at its original bit depth, and to set the Opus encoder at 24 bits (the default - i.e. don't do anything). An explanation of your choice to modify the source audio before encoding would be meaningful. As mentioned, before encoding it's considered best practice to leave an audio source at its original bit depth and samplerate and let the encoder do what it wants to do in terms of resampling or dithering (if any) as it processes the input. The only time this would be inadvisable is if you know the audio will be downsampled badly by the encoder/you know your chosen encoder has a very poor internal/proprietary resampler. That said, no modern (and maintained) open-source resampler library that I can think of "damages" the audio in a way that humans can perceive, so this scenario is extremely unlikely. Certainly the resampler used by Opus is perfectly fine. In fact, the sound science is so well-developed, that I'm quite certain that any "errors" introduced by resamplers have been inaudible for a couple decades at this point and the changes only really become audible at very low samplerates (by design). (I know you didn't resample the source, but I'm still curious about your choice to dither).