Accuracy – Only human captain providers can consistently be 95% accurate or higher. AI captions rarely are this accurate. In addition many CART providers are certified to specific standards. 

Speaker Identification – For one channel audio automated captions have a difficult time identifying different speakers. 

Speed – Top human captioners routinely  hit 250 words per minute and peak over 300 wpm. 

Content control – No matter how good an automated solution is, random words will be generated. This can be entertaining but for many broadcasts this is not acceptable. 

Descriptive content – In many cases a human capitoner will provide cues to actions and sounds that are not part of the transcription. A few examples are music and applause. 

Superior sound processing – Trained humans do an excellent  job of filtering out background sound  like music or multiple speakers and providing accurate captions. Automated solutions will try to make sense out of the noise which generates random text.