A different approach is required if you're mixing for Internet distribution. Here are some things to consider if your mix is intended for the many lossy encoding formats that are required to get your files online, excerpted from the 2nd edition of The Mixing Engineer's Handbook:
- Start with the highest quality audio file possible.
- Filter out the top end at whatever frequency works best (judge by listening.) MP3 has the most difficulty with high frequencies. Cutting these frequencies liberates lots of bits (literally) for encoding the lower and mid frequencies. You trade some top end for better quality in the rest of the spectrum.
- A busy mix can lose punch after encoding. Sparse mixes, like acoustic jazz trios, seem to retain more of the original audio oomph.
- Make sure your level is reasonably hot. Use the “Tips for Hot Levels,” or normalize if you must, but it’s far better to record at a good level in the first place.
- Don’t squander bandwidth. Your song might compress a lot better at 32kHz than at 44.1kHz because the encoding algorithm can concentrate on the more critical midrange.
- Don’t squash everything. Leave some dynamic range so that the encoding algorithm has something to look at.
- Use multiband compression (like a Finalizer) or other dynamic spectral effects sparingly. They just confuse the algorithm.
- Set your encoder for “maximum quality,” which allows it to process for best results. It takes just a little bit longer with today's fast computers, but it’s worth it.
- Remember: MP3 encoding almost always results in the post-encoded material being hotter than the original material. Limit the output of the material intended for MP3 to –1dB instead of the commonly used –.1 or –.2dB.