For starters, don't bother with automatic transcription software, it's still not very accurate, but more importantly, transcribing is an act of analysis, a step that should not be skipped. Use dictation software like Dragon that works well for you and train it (insert requisite How to Train Your Dragon joke here). The training aspect doesn't take too long and it's well worth it for words you use frequently but aren't considered 'standard' by dictation programs. If you don't do this, you end up spelling out these words every time. I'm sure you'll figure it out on the fly, but it's also worth mentioning that you can be more efficient by understanding the limitations of the software. For example, if I say "John Dewey," the program has no issues, but if I say "Dewey," it transcribes, "do we." Thus, it's actually faster for me to always say "John Dewey" while transcribing. The other option is to go hybrid and just type out the words you know are going to be difficult for the program to transcribe whenever you come to them.
The key to transcription speed is the setup pictured below. I happen to have my microphone input and sound output on one split cable, but I anticipate that many people will have separate cables for mic and headphones, which is probably easier to setup. Use a separate recorder (or your phone) and have the audio out go to your headphones. The mic input goes to your computer as it normally would for any dictation task. The other method I've tried out is downloading the interview files and playing them through iTunes and using the play/pause button on my computer.
The idea is to listen to your recorded interviews and simply repeat back aloud what you hear into your microphone for transcription. The significance of this process is that you take on the voices of your interviewees. Every word they say, you repeat it, so the transcribing process serves as a form of analysis. This is not the same as reading a transcript. My experience is that you know your data much better if you transcribe this way. And, it's much faster than typing.
There are a few weird things to get used to such as saying, "Adam colon how did you do that question new line Ryan colon um comma I don't really know period." Dragon translates this as:
Adam: How did you do that?
Ryan: Um, I don't really know.
It's also odd having a conversation with yourself taking on the voices of other people, but you get used to that too. The upside aside from speed is that by the end of the transcription you'll have a strong sense of what the best parts of the interview were as it relates to your investigation.
You may have noticed that I've skirted the issue of 'valid' practices for transcribing entirely and that is on purpose because I think it really depends on the field. I was trained by psychologists who insist on having every um and ah with detailed accounts of the duration of pauses indicated in the transcript. Following the aforementioned method, it is fairly easy to include that information, but it does slow things down when you're counting pauses.