[BUG]: There is a logic problem with volume normalization in Emilia's audio preprocessing

[Hit1ron]

After performing a volume gain, an amplitude normalization is performed again, which will cause the volume gain to fail.

# Normalize volume and limit gain range to between -3 and 3
normalized_audio = audio.apply_gain(min(max(gain, -3), 3))

waveform = np.array(normalized_audio.get_array_of_samples(), dtype=np.float32)
max_amplitude = np.max(np.abs(waveform))
waveform /= max_amplitude  # Normalize

[Hit1ron]