Is it a digital recording device? If so, there’s defintiely options.
The biggest is called Dragon Naturally Speaking. It works very well, but it’s not perfect. You will definitely need to review any transcriptions it produces and you’ll need to pay more attention to enunciation than you’re likely used to. It can definitely hit 95%+ accuracy.
It’s been a few years since I worked with speech recognition, but in my graduate school lab we used CMU Sphinx. There were two broad issues: (1) hardware, as in setting up the mic(s) properly (on linux) and (2) recognition of general speech, which was mitigated by using a restricted, domain-specific corpus. (1) was a standard linux hassle, only incidental to the software and should not pose a problem nowadays. Giving it a whirl will be the only way to see if (2) is an issue; we found that recognition rates were highly dependent on individuals (enunciation, pitch, slurring, etc.).
We used both Sphinx2 (C code) and Sphinx4 (Java); since our software was Java, Sphinx4 was a much better match. According to Wikipedia’s CMU Sphinx page, there’s also a newer version called PocketSphinx (for embedded systems).
Wikipedia’s list of speech recognition software splits it out by operating system. I’m not sure how complete the list is, but perhaps that’ll give you some leads…