Error in mp3 reconversion - docker

[sergiohf]

I have installed ultrasinger in an old laptop, in docker. I have selected to keep cache files. When the job is done, the mp3 file has so much noise that is inaudible, but the wav files (in cache) are perfect. Anyone knows what if this is an issue or problem with config?

[rakuri255]

Is it only with keep cache? Or also without it? Which mp3 files are affected? audio & vocals & instrumental?

Where is the source from? Yt?

It could be related to this: https://github.com/rakuri255/UltraSinger/issues/191

[sergiohf]

Without the cache too. The source was youtube. I followed the process. 1st create the mp3 from the youtube video, then create the wav files with vocals and without vocals. All these audio files are perfect. When all processes finished the mp3 with only vocal and without vocal are with a loud noise.

I dont know if i can share de music files, but if you want i can send to you. PS: Tried in the collab and had the same issue

[Dikakus]

Same issue here. MP3 files sound horrible and noisy.

[NiraNora]

I have the same issue, but with a local file and no docker. The audio gets separated into clean wav files, but the conversion of pydub into mp3 converts them into distorted audios. This is not a mp3 problem, I can do the same ffmpeg conversion that pydub does internally (i.e. ffmpeg -y -f wav -i no_vocals.wav -f mp3 no_vocals.mp3) and the result sounds clean. There must be something else inside pydub that does this. It happens also if I use pydub directly in an interactive shell.