dessa-oss / fake-voice-detection

Using temporal convolution to detect Audio Deepfakes
http://www.atlas.dessa.com
Apache License 2.0
349 stars 86 forks source link

What's the maximum length of the audio array? #9

Open irdance opened 4 years ago

irdance commented 4 years ago

At test/inference time, what's the maximum length of the audio array that can be tested? I'm having trouble finding this out from the code.

If an audio array is longer than this max_length is trimmed?

ranasac19878 commented 4 years ago

Hi Irdance, there is no max length and you can test arbitrary length audio. The trimming only happens based on the threshold dB level which I think is set at 40 dB. So anything less than that value in the beginning or the end of audio will be trimmed but not in the middle of audio. Let me know if you still have doubts.