I have several hundred hours of RealAudio-encoded talk radio shows that i'd like to re-encode to MP3 to play in my car's MP3 CD player.

Half of the shows are encoded with the Real Voice Codec @ 16 kBps/16 kHz/16-bit/Mono.

The rest are encoded (some mono, some stereo) with the Real G2 Codec @ 32 kBps/22 kHz/16-bit.

I've used StreamRipper to re-encode them to MP3, but I keep thinking that there's probably a way to get a higher-quality mp3. I believe StreamRipper is using the Fraunhofer codec.

Any recommendations on how to get the best quality/smallest file size from these already squeezed-to-death files?

Thanks!

I would re-encode them with LAME MP3 at "--alt-preset standard" and let the codec decide the file size. No matter how they are transcoded they will never be any better than they are as RA files. B)

But there's no point in ending up with something at 160 kb/s just to recreate a very lo-fi
recording. Certainly stereo and any higher sampling frequency gets you no quality in transcoding from a mono source.

My first guess is to target a final bit rate at 2x the source bit rate. If you get the same bit rate you will likely get noticeable further degradation. Also, decide whether you care to preserve the 2 channel stereo for those files that are in stereo, or downmix to mono.

Doesn't Lame have some lower quality presets something like "AM radio" or "telephone"?

Choosing the right lowpass will help a lot too. If you are unsure about the lowpass of your files, just open them in CoolEditPro/Audition or a similiar program and look at the spectral view to determine the cut off. Be sure to set the lowpass at the except same value (lame --lowpass xx) when transcoding.

dev0

It wasn't RealAudio, but I was hijacking then encoding 32 kbps/22 kHz mono MP3 "protected" audiobooks as 56 kpbs/22 kHz FhG mono with good results under the recommendations of another here. Did so to save bat life on my iPod (the audiobooks here gigantic files).

Like DonP, I suspect 2x is a good starting point for most of the stuff at this level.

How about using the modified lame compile and something like lame --preset x -a --lowpass y, where x is the double bitrate as suggested by DonP and y is the lowpass as suggested by dev0. (if stereo then skip -a of course)

I have a second suggestion as well when it comes to choosing appropriate setting for mono files. The problem is that there are no tuned settings for mono but my idea is to use --preset x -a and choose x like this:
f=samplingfrequency/2
Then use what I suggested in this post:
http://www.hydrogenaudio.org/forums/index....ndpost&p=143466

For the 16 kBps/16 kHz ra-files, this would lead to a 32 kbps MP3 and for the 32 kBps/22 kHz ra-file it would lead to a 40 kbps MP3 file. However, since we are dealing with transcoding from another format you may want to increase the bitrate a bit to give some extra room for minimizing additional transcoding artefacts.

But honestly I don't have any proof whatsoever that my second suggestion is a good choice, its just an idea I had. If some of the developers have any comments about this, it would be interesting...