Configuration of n_fft and hop_length

[jishengpeng]

I have been experimenting with modifying the hop_length and n_fft parameters in the Vocos framework, but the results have been suboptimal.

Specifically, in the encoder section, I am utilizing the Encodec with a sampling rate of 24000 Hz. I attempted to compress the codec to achieve frame rates of 25 frames per second (fps) and 100 fps, as opposed to the original 75 fps of the Encodec. The specific configurations are as follows:

For 100 fps: hop_length=240, n_fft=960
For 25 fps: hop_length=960, n_fft=3840

Unfortunately, both configurations have yielded very poor results. Could you provide insights into the potential reasons for this issue