All you do is give it your book in any ebook format and it’ll let you select a voice for each character it finds in it. So far I’ve gotten it to work on: Intel mac, linux,windows, and steam deck. For a 5+ hour audiobook (example being the first book in the guardians of Ga’Hoole series), Generating on cpu: 1050 min Generating on 3060 in Ubuntu natively:130 min Generating on windows with 3060: 250 min
You can even set the chapter deliminator which will choose the keyword it uses to detect when there’s a new chapter so the final files are chap1, chap2,… and so on.
The whole thing runs locally for free, been working on this project for quite a while so far, love to hear your thoughts!
that’s pretty cool. it’s neat that it works on different systems and the processing times are quite varied.
running locally and free is a huge plus. curious, what was the toughest part in making it? definitely gonna try it out.
Oh I forgot to ask what OS you have
macOS
Ah good I just haven’t gotten it running apple silicone yet
Thank you! And the top hardest would definitely be a learning the ins and outs of the BOOKNLP output files to extract the info I wanted, and making the python gui,
Your demo audio is fascinating. The narrator voice is great, the Clud(?) voice is awful, the father is pretty good, and the mother is pretty bad. I think I caught that it made a mistake attributing narration to the wrong voice at one point. But overall I like the way it sounds :)
How do you recommend making fixes like “oops, wrong voice here, use the narrator” or “change this voice here?”
Hm I never thought about those cases just suspected no one would ever want to do fine tuning like at, I suppose I’ll have to integrate a way to change it in the gui, at the moment the only way I know how would be to manually modify the book.csv and type in the speaker for that quote your looking at
The book.csv file is located in VoxNovel/working_files/Book/ is just a csv file where each row is a quote from a character/narrator talking, the row “text” contains the quote and the row “Speaker” contains the name of the character talking,
Voices wise tho, it just uses voice cloning from audio samples I gave it of celebrities talking, so you can add your own voices if you want, I just haven’t added a button for it yet, right now you have to go into the tortoise voices directory and manually create a new folder for that voice and put the reference mp3 or mp4 of them talking in it