Because many of these recorded audio chunks are just silence, it's arguably not that useful or interesting to have them on the plot. Additionally they end up taking up a lot of space on the plot that isn't useful. If the audio slices below a certain loudness threshold were removed before doing the PCA and UMAP processing, the dots could be distributed on the plot in a more spread out, useful (, musical?) way.
This thresholding would need to happen in the SuperCollider code that is doing the actual analysis and dimensionality reduction.
For any skeuomorph knob positions that aren't on the plot (because those knob positions produce silence), I propose that the black dot on the plot just go away. We can still hear the playback of that recorded audio! (Even though it would just be silent or near silent).
Because many of these recorded audio chunks are just silence, it's arguably not that useful or interesting to have them on the plot. Additionally they end up taking up a lot of space on the plot that isn't useful. If the audio slices below a certain loudness threshold were removed before doing the PCA and UMAP processing, the dots could be distributed on the plot in a more spread out, useful (, musical?) way.
This thresholding would need to happen in the SuperCollider code that is doing the actual analysis and dimensionality reduction.
For any skeuomorph knob positions that aren't on the plot (because those knob positions produce silence), I propose that the black dot on the plot just go away. We can still hear the playback of that recorded audio! (Even though it would just be silent or near silent).