Text-to-speech technology has come a long way with impressive results benefitting content creators, ebook readers, and visually impaired individuals. Speech-to-text technology offers the opposite conversion and it is beginning to catch up.
Accuracy has long been an issue but algorithms are enabling better results each year. It is good enough for commercial release with some software developers promising as high as 80% accuracy. The text can be edited after processing to correct errors. Even with corrections, this can still be faster than full manual transcription.
The quality of the recording has a major impact on the software conversion. Indeed, even manual transcribers will have a hard time doing their jobs if the background noise overwhelms the voices that they are trying to listen to. If the noise can be minimized, then the output should be more accurate no matter which method of transcription you choose.
This is arguably more vital when creating automated transcript because humans can guess words based on context but programs have yet to achieve this level of sophistication. Record in a quiet environment and perform audio processing if necessary before getting a transcript.