I want to make a wav file using stdout, how can i make? about speaker-diarization HOT 5 CLOSED

Hi!

Glad you like it :)

About your requirement, you want each .wav file to have just the audio content of each speaker? That is:

speaker_1/audio.wav  # A wav file with the concatenation of every speaker_1 part of audio.wav
speaker_2/audio.wav  # A wav file with the concatenation of every speaker_2 part of audio.wav

Is this it? In that case, there's no audio post-processing on this package as-is, so it won't do this automatically, but you could write a python or awk script that takes the output, parses each start-time=xxx, end-time=xxx and speaker=xxx with a regex, and calls ffmpeg to cut the appropriate segment of the original .wav. You can later join all the speaker_x segments together.

This answer is pretty close to what you want, you just need to modify the parsing part to this format: https://unix.stackexchange.com/a/400032/6301

from speaker-diarization.

Is the audio in Korean? The language model that is included is for English language. It would be possible to train a Korean language model with a lot of annotated data and using https://github.com/aalto-speech/AaltoASR but that's quite a big task and I can't guide you through it :(

from speaker-diarization.

Thank You !! < ^ o ^ > !!

I use python and now i'm making split file.

But, i meet a new problem.
I'm a korean, and i test one wav file.
But, performance is not good.

How can i solve this situation you think??

from speaker-diarization.

Thank you. 👍👍👍
I will test other korean .wav file.
And I have to see that source and confirm that i can use it.

Thank you for your rapid answer.

from speaker-diarization.

Closing as it seems this was answered :)

from speaker-diarization.