Speech recognition has come a long way. You can talk your blog posts or even a novel into a microphone. I have used it to dictate 35,000 words of my current manuscript. It’s surprisingly good in my opinion.
Beyond reading the initial training text, I recommend spending some time correcting any recognition errors because the software has to learn your individual pronunciation and word usage. There will be errors, but I’d say 95-98% is correct. With a first draft you have to do a lot of editing anyway.
Even better, you can dictate into a portable recorder and have it transcribed later. You don’t need an expensive one either, just one with low noise (no “shushing”). I find it very useful not to have to sit in front of a PC. I have the small SanDisk Sansa Clip audio player which only costs about $35 and the voice recording quality is excellent. I plug that in, let Dragon transcribe the WAV files, and the text is typed.
The only thing to do is speak very clearly, as though you were a radio/tv news reader. Then it works great.
The techie stuff: Dragon Naturally Speaking by Nuance is the leading software. Get a previous version to save on the price. It’s available for Windows and Mac. On Linux I use Virtualbox to run Windows XP. Transcription from an audio file you recorded might require a premium version. The different versions can be a bit confusing, so read the exact specs before you buy. For example, some versions are multilingual so you can speak your texts in English and Spanish.