I don't think there are any programs that can accurately transcribe pre-recorded audio within acceptable limits. In medical settings, a human transcriber is still very much the norm for documentation (other than providers simply typing their own notes). Far as I know even modern, high-end speech-to-text is still not robust enough to handle this unless the dictator is using the software directly, and even then it requires monitoring and frequent correcting.

Whether paying for software or hiring a transcriber, I think you're going to shell out a similar amount of money.