Hi Jan,
following your line of thought I inspected the silence durations which
as you point out are very precise.
I used ffmpeg to plot the silences and one [ the last ] has a unique value.
once I have this point which is different in each file because of
streaming delay variation, I can trim the audio.
again I used ffmpeg which I think is what sox uses anyway.
the out cue point being human has a little more hysteresis [slop]
I am getting perfect results 100% for the opening and 92% for the closings.
Which is as good as it was when we did these things live.
Thanks for your input.
regards
Robert
On 31/05/21 3:45 am, Jan Stary wrote:
On May 30 09:53:04, jeffares.robert@xxxxxxxxx wrote:
On 30/05/21 4:48 am, Jan Stary wrote:
But perhaps you could use the pips as the silence/nonsilence
cutting point - surely the pips are at very regular intervals.
Exactly what I want to do.
No.
They are 1kHz;
What I propose ignores the pip frequency.
5 of 100ms and one of 500ms
And how long is the silence between them?
Jan
_______________________________________________
Sox-users mailing list
Sox-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/sox-users
--
Robert Jeffares
Communication Consultants
64 Warner Park Avenue
Laingholm
Auckland
New Zealand 0604
+64 9 817 6358
_______________________________________________
Sox-users mailing list
Sox-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/sox-users
