OT: Best Option for Web Delivered Voice Overs

[Joe F.]

I was referring to audio compression/limiting.
An audio codec provides compression/decompression for the file, much  
the same as zip or stuffit compresses files.

Audio compression/limiting works on the level of the audio signal,  
either ahead of recording or on playback.
If your mic is plugged into a hardware compressor, or a software  
compressor is in the signal path, the dynamic range can be manipulated  
to make the minimum and maximum levels fall within a limited range.  
This is what gives you that smooth, professional sound. It also makes  
voices easier to understand and can be used to prevent excessive  
levels that cause distortion.

You want your original "master" recording to be a large file.  
Especially if it's something that needs to be "future-proof".
"Encoding" as mp3 or AAC, or whatever, is compression for file size.
It doesn't make sense to compress an already compressed file like an  
mp3.
You'd want to always be converting from the original "full resolution"  
file for optimal results.

As far as "work flow": it's not clear, but it sounds like you want to  
do the recording within your stack.
If the point is simply to get playback in the stack (recording not  
needed) then it would be far more efficient to use dedicated tools to  
create the recordings, then reference the audio from within your  
project.
If you must record within your stack then you'll have to live with  
some serious limitations.

On Mar 14, 2010, at 3:15 PM, Web Admin Himalayan Academy wrote:

> OK, well no one responded to my actual requirements: fastest audio  
> work flow and as small as possible, I should have added: native  
> RunRev platform... as much as possible...
>
> So I went ahead and did my experiments using Revolution as my  
> recorder. Samson COIU USB Stude condensor USB mic (love it...  
> inexpensive too. on sale now...) I had to learn about the answer  
> record dialog and all the different settings. One opaque point is  
> that if you choose a recording format that does not support the  
> codec you select from the available codec supplied by QT then it  
> usesa default codec and you can't really know what that is so  
> logging exact scenarios/tests is difficult if settings are changing  
> under the hood by QT.
>
> At any rate, I tested many, many combinations. Len's basic point was  
> manifestly confirmed big time. There was really nothing to be gained  
> by using any low compresssion initially. well not exactly. avoid  
> raw... You could record  a slide caption VO -- 8 seconds long using  
> uLaw save to .au  and you get an apparently smaller file at about  
> 400k. Quality quite poor. Record 8 seconds at 64bit floating  44kh  
> sample size 16 to WAV or aiff and you get a big 636K good quality  
> file; but when you run these both thru Switch to MP3, joint, 48 bps   
> you end up with a ~44K file either way... with the result from au  
> being really bad, but the 44k.mp3 that that started out as a big  
> wave file is surprisingly high quality for that size.
>
> So old axiom proved again, garbage in --> garbage out; big diamond  
> in and the little gem that comes out is still a diamond.
>
> So then, and this is the really interesting part:
>
> If I take that 44 K.mp3 file and convert to au with Skitch: codec  
> PCM 8bit; sample rate 6000 channels: mono... I get a 44K.au file  
> that still has great quality that can be imported back into  
> Revolution. Streaming 44K caption VO's should not be a problem even  
> for fairly low bandwidth users. And if we wanted to do this as a  
> revlet and let the stack contain the audio we can do that also.
>
> OM shanti
> Sivakatirswami
>
>