Monday, November 7, 2016

Who said what ?

I'm grateful to radio futurologist James Cridland for pointing me to some new software from Adobe, which, in theory, would allow you to synthesise speech from typed words with someone's 'real' voice.

The programme needs to sample some twenty minutes of your speech to develop a full sound profile, which is then applied over regular speech synthesising trickery.

I have to confess this might have happened earlier. The BBC's R&D department came to me sometime in the last century, to ask if news would be interested in capturing reporter's audio profiles, then using them to rebuild despatches filed down rubbish phone lines, to improve quality. The problem was that, it couldn't happen instantly - and thus wouldn't improve audibility for live broadcasting, two-ways etc. I was also worried about applying the wrong audio profile to a reporter - mixing up, say, Industrial Correspondent Nick Jones with news reporter Val Jones. We didn't take it further.

No comments:

Post a Comment

Other people who read this.......