First off, Logic's time stretching is quite lovely and the best results are when the source audio file is as dry as possible. When time stretching is applied to a file that has FX or Reverb embedded, the results will never be as good, typically.
I keep the time stretching algorithm set to "complex," unless it's for any type of drums/percussion in which case I either "strip silence" to chop it up or use the "percussion" algorithm.
For the best results with time stretching a vocal, your best bet is to use flex time with the flex setting set to "monophonic." I typically do the following:
- Place the audio file in logic.
- Double click the region to open Logic's audio editor.
- Go to file > detect transients.
- I'll then use the +/- in the editor window to either add or subtract unnecessary transients, respectfully.
- Return to the arrange window, select the track and region and activate flex time > monophonic.
- Turn on the metronome (or your tune) and start shifting the transient points (markers) to the proper location. I find the best results are achieved by adjusting transient points that are located at the quietest parts of the wave form as opposed to the middle of a phrase. Also, the less transient markers you use to shift the vocals, the better the result.
Sidenote: You can use the "slicing" flex time option instead of monophonic if you're looking for more a chopped vocal sound a la Danny Byrd.
Also, if the vocal you're editing happens to have been "Acidised," make sure you disable "follow tempo" in the inspector panel. Logic doesn't play nicely with Acidised wav's for whatever reason.
Cheers.