I read about ElevenLabs' AI voice synthesis on Webcurios and was blown away by how good the synthetic voices are. As Matt of Webcurios says, "this stuff is REALLY good now, incorporating natural stresses and pauses and inflections based on punctuation and word-choice and assumed cadence."
Here's the above paragraph read by "Josh." And here's Bella reading it.
You can register for a free account which lets you generate several minutes of narration a month. If you want to make audiobooks, you'll need a pro plan, which will cost a few hundred a month. Paid accounts also let you clone voices by uploading samples of someone's speech. I'd like to clone voices of the 1960s KHJ DJ's and create a never-ending AI version of a KHJ radio broadcast.