Improve vocals detection #5
Description
Need to improve Vocal detection so that we can use it for lyrics extraction from, let's say, Kanye's tracks, which are heavy in music and can cause issues with detecting lyrics for the song.
This workload anyways is only suited bigger Vosk models, but if we can improve the audio filtering and remove most of the noise/music from the audio track, we should be able to achieve the goal.
Don't really know how to achieve it right now, without causing a significant load on the processor. Maybe AVX instructions can help? I'm grasping at straws.
Need someone good with audio engineering to work on this.