Re: [Audacity-devel] Fw: SoX resampling

[Rob S. <aq...@ya...>]

----- Original Message -----

> From: Vaughan Johnson <va...@au...>


>>>  Particularly relevant may be the required spec. for variable-rate 
> resampling, 
>>>  since if libresample falls short of it, it might be easy to fix it—by 
> tweaking 
>>>  its filter parameters for example (this is probably dependent on 
> time-tracking 
>>>  having been fixed though).
> 
> Frankly, I think you're probably most knowledgeable about resampling,
> and could therefore be the best one to review the Audacity code to come
> up with the spec. :-)


:) If fact, I was thinking of this more from the application perspective. For example, Julius Smith (whose algorithm is used by both libresample and libsamplerate) suggests that variable-rate resampling could be used for audio "scrubbing". After googling this term, I found it's a sort of high-speed preview of the audio which I doubt would require very high-quality resampling—even linear interpolation might be okay for this task.

I'm not sure what time-tracks are used for in practice (I don't think they're for scrubbing though); if they're for speeding-up/slowing down a recorded commentary to make it fit scenes/shots in a film/video, linear interpolation is probably out, but I'm sure libresample would give the required quality; however, I suspect that this is a contrived example: more likely one would use a soundtouch-like algorithm here (or just edit the video), to avoid changing the pitch noticeably.  Also, there's the STS/PS effect which covers similar territory.  So I'm a bit perplexed as to what time-tracks are actually used for; googling didn't seem to reveal much, mostly folk mistakenly trying to use them for something other than what they actually do.   This is just for interest, btw; the current plan seems fine :)

Cheers,
Rob