What's The Best Way To Encode Lectures?

[ulTRAX]

I've found some reasonable quality 32kbs lectures on the Internet. Rather than have a mix of 32 and 64kbs files I was trying to convert the 32 files to 64kbs.

My initial impulse was to over-sampling at 44k. Curiously the dbPowerAmp quality bar gave 44k a lower quality rating than 22k sampling. Why would that be?

I tried both settings. Both resulted in files of similar size. So where does the quality issue come in?

Is it because I'm sampling from 11kHz to 20kHz and there's nothing there in the original file? Or because the 44k sampling was compressed 1:11 while the 22k sampling was compressed at 1:5?

One other question... There's a voice setting in the advanced menu. Is there a spectrum chart showing what frequencies this highlights/drops?

Thanks!

[neilthecellist]

Re: What's The Best Way To Encode Lectures?

Converting 32 kbps to 64 kbps: Very bad mistake.

Why? You lose quality transcoding from lossy to lossy. And it doesn't matter what sampling rate you choose either, in fact, raising it will create a scratchy noise that no human will bear comfortably.

I'd recommend that you keep those lectures at their current bitrates. Honestly, there really isn't anything you can do to make them sound any better, quality wise.

[ulTRAX]

Re: What's The Best Way To Encode Lectures?

I'd recommend that you keep those lectures at their current bitrates. Honestly, there really isn't anything you can do to make them sound any better, quality wise.

Thanks for your response. It's been my experience that using graphic equalizers can enhance poor quality sound by emphasizing certain frequencies and diminishing others. So would not encoding using the VOICE feature help in this regard even at the risk of introducing encoding artifacts? So far I haven't be able to discern any difference between the files.

One of the reasons I'm looking to transcode up is perhaps a mistaken belief that if burned onto a data disk... the bigger file might be more resistant to scratches than a more compressed one. If I record lectures from my own sources I do so at 128kbs for this reason even though 64kbs might suffice. Of course I'm not sure if there IS any error correction on a mp3 disk as there if I burn a standard CD. Any thoughts?

[Spoon]

Re: What's The Best Way To Encode Lectures?

Speex is much better at voice if you can use it.

[ulTRAX]

Re: What's The Best Way To Encode Lectures?

Speex is much better at voice if you can use it.

Thanks for responding. I'm new to dbPowerAmp. I don't see Speek as an advanced option in LAME. Is it a plug-in?

[LtData]

Re: What's The Best Way To Encode Lectures?

It is a codec you can download from Codec Central. Link at the top of the page

[ulTRAX]

Re: What's The Best Way To Encode Lectures?

It is a codec you can download from Codec Central. Link at the top of the page

Thanks for that. But I take it it's not compatible with mp3 players. If not it won't serve my purposes.

What do you know about the VOICE setting in LAME?

[LtData]

Re: What's The Best Way To Encode Lectures?

Correct, Speex is a totally different format than mp3.

Absolutely nothing. However, a quick google shows that that preset does an "average bitrate" of 56kbps. I don't know about frequency and channels.

Have you tried converting one or two files with the preset and seeing what they sound like?

[ulTRAX]

Re: What's The Best Way To Encode Lectures?

Have you tried converting one or two files with the preset and seeing what they sound like?

The original 32kbs and the encoded 64kbs (voice) sound identical to me played on a reasonable quality 5.1 system.

But then it's not as if the original has any highs the voice preset can clip. It's sampled at 22k so that gives it a top frequency response of about 10kHz? I would expect to see more of an effect with the voice preset if I used a higher quality source as the highs and lows were de-emphasized.

[Spoon]

Re: What's The Best Way To Encode Lectures?

For any frequency changes enable 'professional frequency conversion' in configuration it is much better.