Wednesday, December 20, 2017


Found via an episode of The World in Words about speech synthesis: Lyrebird. Read a minimum of thirty sentences into your computer’s microphone (example: “I usually like to eat flying tomato salad”), and Lyrebird creates a digital version of your voice.

I tried Lyrebird this afternoon, with just thirty sentences, and the voice that resulted is pretty plausible. (A demonstration.) I could never mistake this voice for my own, but it does sound something like me, a sleepy me, a world-weary me, a me beset by ennui. But Lyrebird doesn’t know how to pronounce ennui, not yet anyway.

I don’t want to begin to imagine the uses that such technology might serve. (That’s me talking.)

comments: 2

Fred Mindlin said...

RadioLab did a piece on this technology recently, pointing out how the intended use of the technology--to make film editing easier when it's not practical to call back an actor--will be immediately overshadowed by its use to manufacture fraudulent quotes from powerful people...

Michael Leddy said...

It must be this episode. My wife mentioned the use of this technology with film editing, and now she remembers that she learned about it from Radiolab.