2.0.7 is out, cut in honour of our own decoders. Press Update in Pinokio to get it.
Our own decoders. YuE2-VAE-RS x3 v1, m-a-p's Standard grown three times and trained by us on music, now decodes every new take by default, with v2 beside it. A take you made earlier can be decoded again with either in seconds, without generating it again. Standard comes back with one tick in Engine → Defaults.
The Narrator, a room for the spoken word: Fish Audio's Drama 3 through audio.cpp, on your card or the processor. Whisper hears every piece and reads again a word lost or added; a take is fixed piece by piece; a long text with [split] is read as a book, part by part in one voice. Built with Fish Audio.
Eight rooms in a new order (Musician, Narrator, Writer, Refiner, Artist, Videographer, Trainer, Librarian), two transcribers (Heretical and Refined), every model download checked three times (pinned commit, size, SHA-256), and an AppImage that carries its own C library.
Next, in 2.0.8: Higgs Audio v3 (Boson AI) as the Narrator's second engine, and the Videographer's own tracks with templates for titles and karaoke.
The full article: https://github.com/igrbible/Ruach_Studio/wiki/Ruach-Studio-2.0.7
The study of our decoders: https://ruachstudio.art/#vae-study
2.0.7 is out, cut in honour of our own decoders. Press Update in Pinokio to get it.
Our own decoders. YuE2-VAE-RS x3 v1, m-a-p's Standard grown three times and trained by us on music, now decodes every new take by default, with v2 beside it. A take you made earlier can be decoded again with either in seconds, without generating it again. Standard comes back with one tick in Engine → Defaults.
The Narrator, a room for the spoken word: Fish Audio's Drama 3 through audio.cpp, on your card or the processor. Whisper hears every piece and reads again a word lost or added; a take is fixed piece by piece; a long text with [split] is read as a book, part by part in one voice. Built with Fish Audio.
Eight rooms in a new order (Musician, Narrator, Writer, Refiner, Artist, Videographer, Trainer, Librarian), two transcribers (Heretical and Refined), every model download checked three times (pinned commit, size, SHA-256), and an AppImage that carries its own C library.
Next, in 2.0.8: Higgs Audio v3 (Boson AI) as the Narrator's second engine, and the Videographer's own tracks with templates for titles and karaoke.
The full article: https://github.com/igrbible/Ruach_Studio/wiki/Ruach-Studio-2.0.7
The study of our decoders: https://ruachstudio.art/#vae-study