Narrators. Who needs 'em? I created a digital copy of my voice on Lyrebird - Create a digital copy of voice by reading like 40 preprogrammed sentences. The end result sounds like me, albeit after 15 cups of coffee and some coke. There's no speed control, but I think they're getting close to nailing it.
I’ve been following this project for a while. The technology is exciting but has a ways to go before the voices are Audible ready. A lot of their samples still sound robotic, but they won’t always.
I tried it again, recorded 50 sentences at a slower pace. Turned out much better. If I went down to the studio to do it, it might sound even cleaner.
Maybe I should try this. People really seem to love my voice, so I could license a digital copy of it to authors needing a cheap narrator for their audiobooks.
Audio samples from "Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions" Google AI Blog: Tacotron 2: Generating Human-like Speech from Text
My experience so far Communication: This company is really bad about responding to emails. I reached out to them on two separate occasions... once when it was far less advanced, and once about a week ago and they didn't respond to either. I use text to speech every single day due to have a reading disability, so the thought of being able to create custom voices is appealing. Business They are sending a lot of mixed message. On one hand they have an entire page devoted to "Ethics" click to another page and they advertise having a "Celebrity voice for your next advertisement," and use Trump and Obama as the voice samples to promote their own business (I'm extremely doubtful that either gave permission). Quality As I said it still has a long way to go until it's ready for audio books. The Obama samples don't sound anything like him, and IMO the only voice that sounded close to natural was "Jane - With inflections" ... At least I think it was called Jane.... the female voice at the bottom of the list. Based on the current quality, I think you'd see A LOT of returned books if you tried put one on Audible without significant sound editing. I'd probably spend at least 30-50 hours in Cool Edit (I'm old school like that), adjusting pitch during dialogue to make characters sound unique and less robotic.
