Silence part

[SunHaoRanCN]

Hi, I implemented audio segmentation with your code and it works great. In the mean time, I also want to generate a list of chunks of silence part which is rejected here. Is there any way I can achieve this?

[yqzhishen]

https://github.com/openvpi/audio-slicer/blob/main/slicer2.py#L76

The list sil_tags represents start frame and end frame indices of all silence parts. You can use the similar method as I get the sound chunks to get the silence chunks.

[EmreOzkose]

To obtain silence parts in seconds, I think below code is enough ?

1. Return sil_tags: (return chunks -> return chunks, sil_tags)
2. Calculate in seconds:

audio, sr = soundfile.read(audio_path)
slicer = Slicer(
    sr=sr,
    threshold=-40,
    min_length=5000,
    min_interval=300,
    hop_size=10,
    max_sil_kept=500
)
chunks, sil_tags = slicer.slice(audio)
for sil_tag in sil_tags:
    start = sil_tag[0] * slicer.hop_size / sr
    end = sil_tag[1] * slicer.hop_size / sr
    print(start, end)

or basically

for sil_tag in sil_tags:
    start = sil_tag[0] / 100
    end = sil_tag[1] / 100
    print(start, end)

[yqzhishen]

@EmreOzkose numbers in sil_tags are frame indices, simply dividing by 100 may not generalize for all hop_size. By the way, there are already some forked version of audio slicer that produces both timestamps and chuncks, like this one: https://github.com/openvpi/SOME/blob/main/utils/slicer2.py