Notions of equivalent metrics
There are two ways to split an MP3:
- Decoding into a wave, splitting, and re-encoding. This is the method Audacity uses and results in lost audio quality.
- Splitting the MP3 directly. You have a lower resolution for where to split, but the result is no loss in audio quality. I've used MP3DirectCut with relatively good success.
Usually you will want to use the 2nd method since it is lossless, unless you really don't care about your audio quality.
Try MP3DirectCut. It's free and works for me.
Specialized MP3 splitters, like mp3DirectCut, cut on frame boundaries. But most MP3s use the bit reservoir to conserve space; a frame's audio data often starts in a prior frame, sometimes 2 or 3 frames back. Thus the frames near the split points are likely unplayable and are silently skipped. If the split occurs in the middle of silence, it's probably of no concern.
If the split occurs in the middle of continuous music, you might notice the skipped frames. To help mitigate this, use mp3packer to expand the mp3 to 320 kbps with minimal bit reservoir usage (-b 320 -r in.mp3 temp.mp3), make your edits in mp3DirectCut, then use mp3packer again to repack (-s -t -z temp.mp3 out.mp3). It's not guaranteed to work because bit reservoir usage might be unavoidable in a high-bitrate file, but it can help, and doesn't result in quality loss.
Alternatively, you can use pcutmp3 to do the split. It preserves the necessary frames and adds gapless playback data (encoder delay & padding info) to a LAME tag at the beginning of the file. Players which support this data will trim the extra samples upon playback.