Software to automatically shorten pauses in recorded speech ???

There is no substitute for doing it long hand.  An editor program like Cool Edit Pro or Adobe Audition makes it easier and way faster than trying to do it in a DAW.  In an editor you can just swipe the unwanted area, hit a key command and it is gone and everything moves down. Simple as that. You can't do that as easy in a DAW as such.
 
 

Kalle Rantaaho
An interesting idea, really, but quite a big order. Like "normalizing breaks" by a certain ratio. I agree to that such a software does most likely not exist.

I am 99.99% certain software list this does exist in the broadcast industry.  They have systems that help them manage their 7-second delay seamlessly, and I think this works on the same principles.  I just don't agree that this should be such a daunting task.  Compared to some of the amazing tings we have seen in VSTs, this seems like child's play.  Basically it would be a noise gate (very well proven technology) combined with an algorithm that simply removes the silence left after processing the noise gate.  There would certainly need to be parameters to control how aggressive the editing is -- or maybe allow the user to specify an overall target length for the entire recording.  But I just don't see that as being particularly difficult compared to other things that people have produced in the universe of digital audio.
 
Now, if we were talking about seamlessly removing video at the same time, that would be a much bigger problem because you would have to analyze the video to see where cuts could be made without being visually distracting, and possible blend together the pieces after the cut.  Again, I think that could be done, but that would be a big order.  Doing just the audio seems pretty simple to me.
 
A more sophisticated version might recognize inflections, such as the cadence and pitch at the end of sentences and leave a little extra space at the end of sentences.  It might also remove "ya know" and other nervous utterances.  I don't expect software to do that today.
 
Doing it by hand is really not a great option.  I'm looking at an hour long radio program from 1988 I'm trying to salvage and I think there is really about 20 minutes of good lively content there.  But I would literally have to remove some space in 300 places.

I am 99.99% certain software list this does exist in the broadcast industry.  They have systems that help them manage their 7-second delay seamlessly, and I think this works on the same principles.  I just don't agree that this should be such a daunting task.

 
You are correct.  This functionality is available in some DAWs.  Sadie for certain which is the software of choice for the BBC, that process is not destructive.  Audition (CS6 and CC), which is found in about 90% of us radio broadcast studios, has a destructive "delete silence" command that works quite well also.  I prefer to do it manually. though, which really doesn't take much more time than listening to the material twice.  

It is unlikely you will find a VST program that can do this for audio.  This kind of editing needs to be done at the program level and VSTs simply cannot do that. Video plugins and scripts are a little more sophisticated in that they can act as external batch processors and alter edits in the timeline.  But they have to be written for each program to be functional.  
 
If you only do it 4,5 times a year, manual is probably your best bet.  Unless you have to do 5-6 per day on a time crunch, you will get better results that way.   

SONAR has a Process > Remove Silence function that is basically a destructive gate that gives you the option to just flatline audio below the gate level or delete it, splitting the remaining audio into separate clips with gaps where the silence was.
 
If you were able to get the gate settings right to have it delete only the significant pauses, you could then increase the tempo of the project to shorten all the pauses by the ratio of the old tempo to the new one (tempo only affects audio clip start times by default). Or you close all the gaps and set them manually using nudge.
 
Nothing is going to fully automate this for you in SONAR, but Remove Silence may help speed the process of doing it manually. Just work with a copy of the clip, and keep applying/undoing with different values until you get the most workable result. Obviously you'll want to err on the side of caution to avoid losing meaningful information.
 
That said. I had to do this a while ago with about an hour of poetry reading and I did it manually, deleting and closing up long pauses by dragging clips, and using volume automation to silence lip smacks and background noise in shorter pauses. It was a chore, but I'm pretty sure no software could have gotten it right without a lot of manual intervention.
 

If a DAW removes silence and creates a whole lot of separate clips then you will have to move them. Also you run the risk of clicks and things happening on the boundaries of the clips themselves. Moving them all is also slower. DAW's are not designed for this sort of work so much. You should have a proper editor anyway, they are very helpful.
 
As I said a proper editor is better. Just swipe the unwanted area, single key command and its gone and all material moved down in one hit. Faster. Also you start to get good at identifying things like grunts and breaths and things too. You do have to make sure the waveform is sufficiently high enough as well so you can see the low level stuff a bit easier.  All these edits are non destructive too so if you accidently remove something important it is very easy to undo and go back to where you were a few seconds ago and re do something.
 
It does not take very long to do even an hour of material. You should be able to do about an hour program edit in about two hours.

cparmerlee

brundlefly
SONAR has a Process > Remove Silence function that is basically a destructive gate that gives you the option to just flatline audio below the gate level or delete it, splitting the remaining audio into separate clips with gaps where the silence was. ... then increase the tempo of the project to shorten all the pauses by the ratio of the old tempo to the new one

Interesting idea.  I wasn't really thinking about doing this under SONAR, but I'll give that process a try.  Even if it isn't the result I need, I'll learn something in the process.  Thanks.

Never tried but I believe (free) audacity does this - http://manual.audacitytea.org/man/Truncate_Silence

cparmerlee

Brando
Never tried but I believe (free) audacity does this - http://manual.audacitytea.org/man/Truncate_Silence

Thanks for that pointer.  I never noticed that in Audacity.  It seems like the parameters should give me enough control to make this sound natural.  With audio, I don't think it should be a big problem.  Video would be a much bigger problem making it appear seamless.
 
On edit: I just tried this and it works GREAT !!!.  In under a minute I was able to remove 5 minutes from a 22-minute interview (not counting the part where they played music tracks) and it sounds perfectly natural.  It actually is easier to listen to than the original because I found some of the pauses annoyingly long.  I adjusted the Audacity defaults a little.  I set the silence threshold to -30 dB, space compression to 5:1 and the max pause to 900 ms.  I'll go back through and remove the grunts and other extraneous sounds, but that will only take 20 minutes.

Glad it worked. Definitely something I will keep in mind in the future when doing this type of thing myself.

bitflipper
The best approach, I think, would be a hybrid process: remove the largest pauses with Audacity and then tidy it up manually from there.

Yes, I agree.  That's what I did on this project.  Audacity's tool is an analogous to a dynamics compressor.  It doesn't automatically truncate all silence down to a set value.  If you are over the minimum value, it reduced the gap by a ratio (4:1 by default).  If you are still over the max gap size, it truncates down to the max size.  So with a ratio of 3:1 or 2:1, the original cadence is preserved (in relative terms).  In my case the interview was from 1988 with a couple of older guys who were slow-speakers to start with.  In 2014, this pace is excruciating, so I wanted to change the cadence.
 
I probably did another 40 edits manually.  Some were to further tweak the cadence.  Others were to eliminate stutters.  And in some cases, there was enough background noise to throw off the Audacity tool, leaving 2- or 3-second gaps that I had to remove manually.
 
I think there are "zero crossing" settings in Audacity that can help eliminate pops at the edit points.  But I didn't do that properly and ended up with a couple dozen light pops.  I simply ran this through Izotope RX3 to eliminate those editing artifacts.
 
After speeding up the cadence and clearing out most of the time spent playing audio tracks, what was originally a 50-minute recording is now a higher-impact 20-minute program.