This is a most excellent place for technology news and articles.
I’m curious enough to try this.
Audiobooks are made or broken by the quality of the narration. It’s not enough to simply narrate written word, the narrator has to “act” as well. Given the quality of other arts engaged by AI I’m going into this skeptical.
Love all works narrated by Kramer, also all the newest Discworld recordings (absolute masterpieces)
I actually just tested this out with a snippet from Small Gods. It still sounds very robotic and, in my opinion, is nowhere near good enough for these books.
The timing between sentences and paragraphs is inconsistent, the intonation for sentences with ellipses is completely wrong, hyphenated sentences are treated as continuous strings, and it can't get names of characters or places remotely close to their intended pronunciation.
It might be fine for reading a news article, but it is nowhere near ready to perform an audiobook.
I never thought there was anything wrong with the first narrations. I hope those aren’t gone for good.
I've never listened to the originals, but I love that 41 books (I think, I'm on eleven now) have been read by the same actors (including Bill Neighy) throughout. Gives a great feeling of continuity. Every time Death speaks without the book hinting I think, "Oh something bad is happening to somebody."
I think the same. I don't think "Ai" would be able to give a correct intonation for each situations
I won't even try. It's not something I want machine to do, I want someone's opinion to go into how a book is narrated.
I had somewhere Asimov's Foundation audiobooks, in Russian, recorded somewhere around 1991, and I don't think a machine can do that.
You should give Ender's Game a listen, the one narrated by Harlan Ellison and Stefan Rudnicki (I borrowed from my local library). The performance is excellent.
I was actually searching for this exact idea like two weeks ago, and came up empty handed. Sometimes things do work out, I guess.
For the unaware, there’s an app called Prologue, which allows you to listen to audiobooks on a Plex server. Plex notably doesn’t have native audiobook support… The first-party PlexAmp app doesn’t support chapters, for instance. You can have the library remember your listening progress, but that doesn’t help much with a book audio file is 12 hours long. As a workaround, Prologue uses Plex’s service to access the files, then is able to read the chapter data from the m4a/m4b file.
Might try this. I've used TTS Server which uses the MS Edge Api to convert text over the internet to audio with few problems. Almost sounds natural except for the occasional word... Would love something better.
Edit: dang, PC, not Android. Anyone know of a way to run this mobile?
Interesting. Just a couple of years ago AI generated audiobooks were unlistenable for me due to the dead voice delivery but the example here actually sounds like it would be alright to listen to.
Definitely want to give this a try for some books that have never had an audio book recording. Thanks for posting!
Kinda off topic but does anyone know a reliable thing that does the opposite, i.e. a good speech to text tool? There's a few audiobook series that I'd love to read but are only available in audio format
I thought about creating something like this some time ago, guess I should think less and do more
Sweet. I'll see if I can get working later on. Maybe I missed it, how much RAM is required for this?
I'll give it a shot.
I tried something similar a year or so ago with my own script. What I found was, the longer the book, the harder the ai has on having the same voice.
I eventually got around this limitation by using piper. https://github.com/rhasspy/piper
It was good enough for my purposes. But it definitely sounded mechanical.
Let me know if it's worth the time.
For sure, just give me a nudge as I may forget to come back haha. I'll try it out tonight if I get a chance.
Been hoping for something like this. Hello waiting ebooks!
