@TimToplak -
In the spectrogram view, this project seems to use a fixed scale for the frequency spectrum. Sometimes the most interesting bits of sound are limited to 0-6k Hertz, so it would be useful to "zoom in".
We tried feeding it files with sampling rates at 8k and 16k hoping that it would scale to a lower Nyquist, but that does not seem to work. It created a model that was completely flat in the upper end of the spectrum. That tells me that the range is static somewhere in the code. When I dug into the code:
I know audio, but very little about WebAudio API. Would it be possible to get the sampling rate from the upload file? This should be a simple way to force the model to have a more limited frequency spectrum. Users could simply down-sample th file before uploading to your project.
A more complex solution would be some additional controls to zoom into the lower end of the frequency spectrum.
Thanks for your consideration and THANKS for sharing this project!
@nwolek
@TimToplak -
In the spectrogram view, this project seems to use a fixed scale for the frequency spectrum. Sometimes the most interesting bits of sound are limited to 0-6k Hertz, so it would be useful to "zoom in".
We tried feeding it files with sampling rates at 8k and 16k hoping that it would scale to a lower Nyquist, but that does not seem to work. It created a model that was completely flat in the upper end of the spectrum. That tells me that the range is static somewhere in the code. When I dug into the code:
I know audio, but very little about WebAudio API. Would it be possible to get the sampling rate from the upload file? This should be a simple way to force the model to have a more limited frequency spectrum. Users could simply down-sample th file before uploading to your project.
A more complex solution would be some additional controls to zoom into the lower end of the frequency spectrum.
Thanks for your consideration and THANKS for sharing this project!
@nwolek