Speech recognition software such as Dragon NaturallySpeaking and CMU Sphinx have automatic audio-to-text transcription features. Dragon is proprietary software, while Sphinx is free and open source. Online cloud transcription services, including Transcriberly, also offer machine-based audio transcription for a fee.
As of 2015, automatic transcription programs lack the accuracy and reliability that humans provide, although most operate faster than human transcribers. Transcriberly claim a 65 to 95 percent accuracy rating for the machine service it offers, compared to a 98 percent accuracy for its human transcription services. The reason for this difference is that current speech recognition algorithms have various difficulties with homophones, word boundaries, heavy accents and background noise.