As far as I can see, the first silence in the selection (whether white space, generated silence or sufficiently low level noise) is always extended, but the final silence (defined as above) is not extended.
Gale
Leading silence is extended if there is any leading silence.
There is alway trailing whitespace at the end of all tracks.
Is that not a gap?
If you have a recording with 40 or 50 silences (gaps) that you want extending, then it makes a lot of sense to automate the process if possible.
The user may also want silence added at the end, or they may not. I would guess that in the majority of cases, adding space at the end would be unnecessary. In your example of playing multiple files one after the other, we still don’t know that the user will want a gap there, and if they do, whether they want that gap to be extended by the same amount as other gaps.
We can make plug-ins as complex as we want.
We have the following design choices:
Add silence at start: Always / Never / Only if leading silence / User option.
Add silence at end: Always / Never / User option.
Against the initial “design brief” the options that I chose were:
Add silence at start: Only if leading silence.
Add silence at end: Never.
As Robert points out, if we extend trailing silence, then selected trailing white space will be rendered as silence and extended.
Unfortunately Robert’s suggestion of removing leading and trailing white space may not always work as expected. If the audio has been ripped (digitally extracted) from an audio CD, it is quite possible that the (non-white-space silence) is absolute silence, which is indistinguishable for Nyquist plug-ins from white space.
Given that it takes 2 seconds to add (or delete) trailing silence manually, is it worth the added complication of an additional control? We have previously run foul of “special casing” trailing silence in Sound/Silence Finder, which encourages me to keep it simple in this effect and either always or never extend trailing silence. Which is more likely to be useful for users?
A dentist would probably say so, even if it is the last tooth in the row that is missing.
From an audio book perspective, we have initial silence and trailing silence and gaps between words.
The leading and triling silences are m well-defined and need rather manual changing to have an absolute length.
I can’t exactly follow. What is the problematic point in this?
As I said, the white space is re-added at the end of the process.
If the CD is ripped as one track, the initial silence would stay the same length (as long as it is not above 0), the gaps between songs are extended and the end is removed (where it is zero, fade-outs are preserved)
Nyquist plug-ins cannot produce white space. The best they can do is to create silence.
Example:
Track 1 is 10 minutes duration, starting and ending with sounds, with silences between.
Track 2 is 5 minutes duration, starting with a sound, with multiple silences, and 2 seconds of “almost silence” at the end.
Track 3 is 3 minutes duration, starting with 10 seconds of absolute silence, and 10 seconds of “absolute silence” at the end.
Select all, then extend silences by 5 seconds:
If we extend the final silence or near silence, then we end up with:
Track 1 has no silence added at the end.
Track 2 has 5 minutes and 7 seconds of silence at the end.
Track 3 has 7 minutes and 10 seconds silence at the end.
If we strip out zero value samples from start and end.
Track 1 has no silence added at the end.
Track 2 has 5 minutes and 7 seconds of silence at the end.
Track 3 has either, 5 or 15 seconds of silence at the beginning (depending on whether we replace the stripped out absolute silence) and either 5 seconds, or 7 minutes and 15 seconds of silence at the end (depending on whether we replace the stripped out absolute silence).
What “should” happen if we are extending all silences (but can’t until version 4 plug-ins):
Track 1 has no silence added at the end.
Track 2 has 7 seconds of silence at the end.
Track 3 has 15 seconds of silence at the end.
What “does” happen now:
Silence is not added to the end of any of the tracks.
Track 1 still has no silence at the end.
Track 2 still has 2 seconds of (almost) silence at the end.
Track 3 still has 10 seconds of “absolute silence” at the end.
As I see it:
Nyquist gets three tracks, all 10 minutes long.
First, the sound is trimmed to exclude absolute silence.
This does not affect the first track because it is the longest and has sound on both ends.
The initial silence of 10 s for the third track should be remembered since it could be white space and it has to be added after processing.
The final silence of the same track is eliminated in the end.
The 2 seconds of the second track are so far not removed since it is e.g dithering noise. It will be extended by 5 s (hopefully with similar content)
However, you can ignore the first and last, non-absolute silences if you wanna do so.
If that’s the rule, so be it. I’m just saying that a mere user has no idea what the behaviour will be until (s)he tries the plug-in. So far I’ve been puzzled by the behaviour, though I think not extending trailing silence is more reasonable than extending leading silence.
It doesn’t see to me like a gap within the audio content (the content starts after the leading silence).
Gale
So you’re saying that you want the leading silence to not extend and the trailing silence to be extended?
or that you want all silences to be extended?
“Want” is too strong a word, but my incorrect assumption was that the default would be for all silences to be extended (because there is nothing to suggest otherwise).
If you don’t want an option to control this, to me the most logical behaviour for the use case you gave of extending gaps within audio would be for neither leading or trailing silence to extend. Would that create a problem if applied to multiple tracks?
However if we are considering applying to multiple tracks it seems to me the use case may go beyond inserting space for narration and if so we actually want some options to control what happens to leading/trailing space.
Gale
I agree that is a very reasonable assumption (thanks for the feedback). In fact I’d go further to say that the name “Extend Silence” strongly suggests that assumption. The only reason that the plug-in does not currently do that is because it was written for a specific use case. As a “release” plug-in I think that extending all selected silences would be best (most intuitive) as the default.
I may provide “Ignore trailing silence” (suggestions for a better label?) as an option, as this would be very useful (given current limitations of Nyquist plug-ins) for the not-unlikely use of applying to multiple tracks.
There are some features described HERE that I quite like, so I’ll see how complicated the interface looks in the final version before deciding on the “Ignore trailing silence” option.
Doh! That’s not going to work for the very reason that I said to Robert (just a few posts back). Nyquist can’t tell the difference between absolute silence and white space, so there is no completely satisfactory solution to the issue of trailing white-space within the selection. We would either need to trim off trailing zeros, which would be wrong if the audio has trailing absolute silence, or trailing white space will be rendered as silence. Probably best not to bother with special casing the final silence and just extend all silences and selected white space (that meets the silence detection criteria).
Topic split. The latest version of this plug-in can be found here: https://forum.audacityteam.org/t/extend-silences/34612/1