Voice to text is a demanding area in the COVID crisis. People are staying in home but work is going on in some sectors. If I consider the work of a project manager, he says he needs to give a lot of instructions to his stakeholders. There are various options for executing his instructions. But among all voice to text is the best method. He does not need to write any mail not even need to give a phone call. The amazing thing is that Google’s voice to text service is excellent which is driven by AI.
Matt Reed’s little digital fun focuses on one of the most exciting technological developments of the day. How long will it take before the software can speak to us so well that we no longer notice that no one is on the other end of the line? The memories wake up from the first presentation of Google Duplex two years ago. The online giant’s software called a hairdressing salon by phone and made an appointment for a customer. The staff member in the hairdressing salon – it was never clarified whether this was really a human being – at that moment did not seem to be able to distinguish whether the caller was human or machine. And that didn’t matter for the exchange of information, namely the appointment. Nevertheless, the outcry was great. The techies gathered at Google’s IO conference screamed with enthusiasm. The media, consumer advocates, and ethics councilors screamed in horror at the “fraud” that the machine committed on the poor hairdresser.
The main criticism was that the Google software added: “irrational” elements to the speech, namely hesitant “Hmms” and approving “Ahaas”. “It’s crazy,” says Sam Crowther from A Million Ads. “For decades we sound engineers worked our fingers sore to remove ‘Hmms’ from voice recordings to get the better sound. And today automated speech production is characterized by the fact that the result sounds more human through such ‘Hmms’. ” This also applies to the spectacular “JFK Unsilenced” project by Cereproc. In years of detailed work, the sound engineers collected snippets from speeches by former US President John F. Kennedy and “reconstructed” a speech that JFK never gave. He should have kept her in Dallas on November 22, 1963, but was shot on the way there. The artificial speech not only contains the authentic voice of the President but is also filled with crackling and atmospheric noises that are supposed to simulate a “historical” recording. In fact, the developers overshot the goal.
Speech to text is really a breakthrough for work at home. People can think more deeply and can give accurate instructions. Recipients can complete the job easily and can perform effective delivery of the job. It was really a demanding phenomenon. Its a history.