Re: [Audacity-devel] Pitch Shift

[Steve t. F. <ste...@gm...>]

On 5 March 2016 at 19:31, Bill Unruh <un...@ph...> wrote:
> On Sat, 5 Mar 2016, Roger Dannenberg wrote:
>
>> In my view, effects are what they are. One should not assume an effect is
>> an
>> imperfect implementation of some "ideal" that is well-understood. Time
>> stretch is a
>
>
> Well, I think that you are overstating this. When someone says pitch shift,
> what they have in mind usually is not "pitch shift with lots of distortions
> which make it sound completely unlike the original" (Yes, I am overstating
> to
> make a point). While some may like the latter, I think that there IS at
> least
> a rough "ideal" that people have when you say "pitch shift" to them. That is
> not to say that the above may not be useful to some/many (Just as some might
> enjoy the sound of the announcements in the London tube stations in their
> only
> tenuous relation to human speech).
>
>> good example: the "paul stretch" effect had a lot of support, not because
>> it was
>> free of artifacts but I suggest because of the artifacts. Composers and
>> sound
>> designers are often interested in artifacts and failure modes more than
>> "clean"
>> effects. We should not treat effects like fragile eggs that must be
>> handled
>> carefully. If someone wants to throw them at the wall or make an omelet,
>> we should
>> not stand in the way. Of course, we also do not want novices to be
>> confronted by
>> unexpected results or be misled into experimental realms they have no
>> intention of
>> exploring, so careful design is needed. -Roger
>
>
> I do agree with these last sentences, but that should also extend to the
> names
> given to the effects. If it is possible to make a pitch shift which
> preserves as many aspects of the
> original sounds as possible, other than the pitch, I think it is a goal
> worth
> going for.

The "quality" of these time-stretch effects is highly dependent on the
character of the audio being processed. On pure sine waves, the
quality is very good even at extreme settings, but, for example, for
piano music, the percussive nature of the sound can become noticeably
echoey even with relatively small amounts of stretch.

The "SoundTouch" library used by these effects provides for tweaking a
number of parameters that can provide significant improvement to the
sound quality, but the "cost" is that adding more control increases
the complexity of the UI. One of the design decisions for this type of
effect is in striking a balance between how much control is offered to
the user, with the complexity of the interface. Many of our users have
little experience with audio software and are not very technical. For
those that can manage to use a more complex interface, we have the
"Sliding Time Scale / Pitch Shift" effect.

At some point I am hoping to add the SBSMS algorithm as an option in
the "simple" Change Pitch / Change Tempo effects, though this may not
be necessary after M.C.Sharma's improvements to these effects.

Steve


>>
>>
>> On 3/4/16 12:45 AM, M C Sharma wrote:
>>       Dear Roger and other friends,
>>
>>       I did go to the link mentioned by Mr.Steve regarding pitch shift and
>>       change tempo effects. Unfortunately I did not find the code which
>>       actually does the pitch shift etc.. Any way ,I would like to clarify
>> few
>>       things:
>>
>>       1. If the work is based of WSOLA , then as per the original paper
>> the
>>       algorithm was in the context of speech.I may be wrong but this
>> algorithm
>>       will simply not work well for polyphonic sounds.
>>
>>       For monophonic sounds (speech / single instruments) even if the
>> simple
>>       pitch determination based algorithm is applied with a proper scheme
>> it
>>       can stretch sounds to 4 times it's duration and so also the pitch
>> can be
>>       made to attain 4 times it's value without any distortion at all.
>>
>>       Incidentally in the current effects even the speech is also
>> distorted
>>       very badly if you reduce tempo below -50 % or if you raise the pitch
>> to
>>       twice it's value ??.
>>
>>       2. In my view both "change tempo and change pitch" effects can be
>>       implemented flawlessly in frequency domain of course within
>> reasonable
>>       limits. (.25 to 4 times).
>>
>>       3.As suggested by you the new function in Nychist has all these
>>       capabilities hence the problem is likely to be resolved in future.
>>
>>       4.We should be careful about assigning limits.Giving astronomical
>> limits
>>       can be counter productive.
>>
>>
>>
>>       Regards
>>
>>       [signatureline.htm@Middle]
>>       Get your own FREE website, FREE domain & FREE mobile app with
>> Company
>>       email.
>> Know More >
>>
>>
>
> ------------------------------------------------------------------------------
>
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>