I use the OpenCL version from the latest release to separate the vocals from two songs last night, I use the GPU acceleration with ensemble mode with all 3 models available on the OpenCL version. I first had it output in .WAV format, then ran it again in .MP3 format. It deleted the first output in .WAV format when it made the .MP3 format files. I was using an AMD RX 6700 XT GPU, and I did not notice any error messages. The original song files were .MP3 from a Youtube downloader service. (https://www.youtube.com/watch?v=5CeX5VEo10c and https://www.youtube.com/watch?v=abzbVFuxigg) Each time I ran the program separating the vocals from the instrumental, it took around 5 minutes for both files together.
I use the OpenCL version from the latest release to separate the vocals from two songs last night, I use the GPU acceleration with ensemble mode with all 3 models available on the OpenCL version. I first had it output in .WAV format, then ran it again in .MP3 format. It deleted the first output in .WAV format when it made the .MP3 format files. I was using an AMD RX 6700 XT GPU, and I did not notice any error messages. The original song files were .MP3 from a Youtube downloader service. (https://www.youtube.com/watch?v=5CeX5VEo10c and https://www.youtube.com/watch?v=abzbVFuxigg) Each time I ran the program separating the vocals from the instrumental, it took around 5 minutes for both files together.