I have two MP3 tracks from a live recording. The transition between them is more abrupt than I'd like. I want to cross-fade them and then re-cut the two tracks. The sound editing software I have will let me audition the transition and cut and save the files, but in doing so it re-encodes its WAV workfile into MP3 and that is going to degrade the sound. Is there a Windows application that will allow me to do this natively in MP3 format so there is no transcoding, or maybe just transcoding of the cross-faded frames? Thanks.

You can fade in/out an mp3 without reencoding but I don't believe it is possible to cross-fade. Reencoding just the cross-fade part may be possible but tricky. I don't know an application that will do this for you.

It's eminently possible, in theory. You have to transcode the cross-faded part of course, and handle the bit reservoir correctly when splicing into the other MP3 frames which are copied verbatim. Our software can do it, but not in a way that would be useful to you I fancy as the functionality is wrapped up in a vinyl-to-cd package designed for non-technical users and the way we import existing files is a bit clunky currently. I don't know of another program that does it, sorry.

But maybe you are worrying over nothing. If you re-encode at a high enough bitrate I doubt that you will hear the difference.

Paul Sanders
http://www.alpinesoft.co.uk

Umm, sorry, but i fail to see how transcoding the crossfaded part reliable is possible. When transcoding normal files which have seamless trackchanges, pops can easily appear, which is why its never a good idea to transcode tracks with seamless trackchanges, especially in the case of MP3. Wouldn't the same problem appear if one wanted to transcode a section of a track?

Discontinuities can happen even without transcoding, but I don't see how splicing a transcoded section of cross-faded audio can't result in an improvement over the OP's current situation.

You can transcode anything, but as Apesbrain points out, you lose a little quality each time you do so. The splicing is the tricky part, and there are two areas of difficulty. One is the bit reservoir, where frame y can refer back to frame x for part of its encoded stream of bytes (i.e. some of frame y's bytes are actually in frame x, ). We solve this problem by repacking the frames around the splice.

The other problem is that the output of both the MP3 encoder and decoder is influenced by the last one or two input frames, not just the current one. We solve this one by giving the encoder and decoder a bit of a 'run up'. To get the splice exactly right, we also have to allow for a 48 sample delay (plus a couple of silent 'preamble' frames) introduced by the LAME encoder. I know it all sounds a bit dubious but it does seem to work.

Well, if anyone is still awake out there, now you know, and perhaps the complexity of the process might explain why software that does this is hard to find. But you might find you can use a fader + a stitcher and get decent results. I'm not sure what tools there are out there but there must be something.

First, thanks for all the replies. You're right that the difficulty lies with "stitching" the file back together after the middle cross-faded section has been transcoded. I've tried a few applications so far (musiCutter, Music File Merger), but none are able to do it silently. Reading the earlier posts about the potential of MP3 frames referencing earlier frames for data, I can understand how this might be a problem.

I did try transcoding the entirety of the two files, but the originals aren't that great to begin with and even at LAME V2 new artifacts were introduced.

As I probably won't have need to copy these files again, I may just save the "corrected" WAV files as FLAC and avoid the transcoding issue entirely.