also if audio file same quality as device native-sampling rate, it would not have to resample again

also android resamplers use up battery and may be of low quality.. detect native sampling rate maybe by playing a sine wave chirp at various sample-rates and looking at the spectrogram

roughness … Reduces the amplitude of alternate waveform cycles in order to make the voice sound creaky.

if's it's artificial & not anti-aliased , if not adequately over-sampled introduces artifacts/distortions into the audio

[espeak-ng:master] New Issue Created by cmrdt:
#178 upgrade internal audio quality (sample-rate & bit-depth)

also in espeakedit (not ng) the resampling quality seemed to be so-so (maybe not sox's fault, maybe command-line-settings) but resampling from a low-quality source-wav file isn't the same as starting from one of higher quality (lower rates lack higher frequencies)

default currently is 11025×2 Hz 16-bits maybe increase to (18×2)kHz+ and 32-bit-float as compile option; & rather do downsampling from higher quality if need be (update syllables containing high frequencies: s f z ʃ ʒ t k θ ð v d ) aka syllables that people (old or hearing loss) may have trouble distinguishing

and if using sox use best-quality sinc option, max precision