Speech speed compression... ever had to do it?

[rasputin1963]

I'm curious: I may soon be making a 3D animation for a local production house.

 

As is typical with this artform, you record the soundtrack first, then animate your characters to the voice.

 

Another convention of animation, is that they (usually) have to be fast-paced; for some reason, 3D anims taken at a normal human speed seem sluggish.

 

So I shall probably have to do some judicious "speed compression" of the speech soundtrack before I animate... where the speech is speeded up ever-so-slightly (or sometimes very dramatically, as with those strange financial disclaimers you've heard at the tail-end of auto commercials on the radio), but of course the speech mustn't sound like chipmunks.

 

Have any of you ever done this kind of speech compression (for any purpose)? How did you go about doing it?

[Synonym Music]

??? Standard in Pro Tools (or pretty much any DAW while we're at it) these days. Pretty good algorithms up to a 10 BPM difference; slowing down sounds nastier, quicker though... In fact you should be able to really push the limit.

 

Before time-stretching it was acceptable pitch variation (Varispeed) with no more than a semitone of difference, then per sample phasing which didn't speed up individual words but did shorten or lengthen the time between them. Complex techniques include both of these.

 

I can pretty much guarantee you those disclaimers are sped up in Pro Tools or somesuch. It's certainly not hard to do.

[Synonym Music]

*sigh*

 

Ok fine. Give me a script and I'll at least try.

[Lee Knight]

I use 3rd party plugins, the Pro Tools freebie and...

 

...even the cheapo in Sound Forge. I record a lot of nonprofessional voice "talent" for telephony purposes.. Car Dealer owners, etc. They call into our voice mail system and I'll retrieve their reading and I do my best to make it presentable. One of the shortcomings of nonpros is "hearing the read." They pause for a second as they mentally grasp the copy. The end result is a vowel that's stretched out unnaturally to attempt to cover up the fact that they're reading. Simple fix...

 

Select the elongated phrase or vowel if there's a clean enough break and adjust the speed by 85% initially. That usually makes them sound like they know what they're doing. Note this is only a phrase within the entire script. Maybe several different phrases, but not the whole thing.

 

So yeah. I do it a lot. Sometimes with finesse, sometimes with a sledgehammer.

[veracohr]

I used to have to fit radio spots and recorded programs into a time frame. There was a nifty tool in Cool Edit Pro, which we used, that just deleted silent portions of the audio file. You could set tolerances for threshold level and length of silence, and unless you were doing drastic time compression it was fairly unnoticeable. I'm sure there's a similar feature on other audio editors. That's probably how most of those super-fast disclaimers at the end of radio ads are made.

[Rudolf von Hagenwil]

In most digital audio production software today, you simply drag one end of the stem to the required lenght and the audio plays the stretched or shrinked time with or without detuning and non-destructive.

[Hard Truth]

I do video editing and occassionally time compress some of the dialogue or narration slightly to match the duration of the cutaway visuals. Before going that route, I recommend editing the voice work to remove any excessive pauses. When I edit an interview I often remove many of the umms and errs, repeats of phrases, pauses and other things people do to pause to think while speaking. That usually makes time compression unnecessary. A good rule of thumb is to reduce the time left by removal of a pause of umm to 66% of its original length.

 

If its not too late, it would probably be best to record a take with the actors speaking at a faster pace, which might save you lots of editing time and sound more natural. If you watch an old movie you'll notice that the actors often speak unnaturally fast, but it usually seems OK in context.