Hi Jesper,
I think if you use clean samples - ie person speaking close to the mic
(but not clipped), you might be able to make it a bit better.
Increasing the bit rate is a major job - but you could try other open
source codecs like Opus.
- David
On 18/12/20 8:39 am, Jesper Norberg wrote:
Hi David,
So, these samples I picked up from drill sergeants on youtube, the
ones I use in the end will be cleaner.
Ideally I would like to get at least some amount of agitation in the
voice to work, it does seem like a very plausible scenario if we're
talking soldier communication potentially mid-combat.
I'm going to experiment a bit to see if I can figure out a mastering
chain that filters out what confuses the algorithm.
Do you think it's possible to improve the results by increasing the
bitrate? If so, how would you suggest I approach it? Is it sufficient
to expand on the 3200 algorithm or would the assumptions you mentioned
also need to be changed?
Best regards
Jesper
On Thu, Dec 17, 2020 at 8:18 PM David Rowe <[email protected]
<mailto:[email protected]>> wrote:
Hi Jesper,
Interesting samples! It seems to do OK on what I would call
"clean speech" - e.g. the female starting at 1:02, and male at
1:30, but struggles with the people shouting, or samples that were
recorded from microphones far away from the speaker (room
acoustics). I had trouble understanding some of the source samples.
I'm sorry but I don't think there is much we can do with
pre-processing to improve the quality. To get the very low bit
rates we make certain assumptions about the input speech, e.g. the
people speaking close to the microphone.
Cheers,
David
On 17/12/20 10:24 pm, Jesper Norberg wrote:
Hi everyone!
I'm working on a project where I want to add a digital radio feel
to military voice lines. I found codec2 and really like the sound
in the examples, but I'm having trouble reproducing the same
audio quality. Some of it sounds good, but especially when the
voice line gets more agitated it starts breaking apart. Appending
an audio sample to show what I mean, this is a 8khz 16 bit mono
sample with a codec2 bitrate of 3200, which from what I
understand should be the highest quality. The sample is first
women then men, both with examples in descending intensity.
I'm a bit new at this, is there any preprocessing/settings that I
could apply in order to improve the quality?
Original
https://drive.google.com/file/d/1IWyV_CwK0KYXC0_ZN3t3QJNk9UCwTOp9/view?usp=sharing
codec2 - 3200
https://drive.google.com/file/d/1Ww1QsNUw8s5TLQXQnBln4CkgpmjUuVFS/view?usp=sharing
Best regards
Jesper
_______________________________________________
Freetel-codec2 mailing list
[email protected]
<mailto:[email protected]>
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
<mailto:[email protected]>
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
