Ruach Studio 2.0.7: our own decoders, the Narrator and eight rooms

[igrbible]

Ruach Studio 2.0.7 is out: the release. We cut it in honour of our own decoders, five days after 2.0.6.

The full article is in the Wiki: Ruach Studio 2.0.7: our own decoders, the Narrator and eight rooms. Here is the short of it.

• Our own decoders. YuE2-VAE-RS x3 v1, m-a-p's Standard grown three times and trained by us on music, decodes every new take by default; v2 stands beside it. Both run in true float32 on NVIDIA cards, and a take you made earlier can be decoded again with either in seconds. Standard comes back with one tick in Engine → Defaults. The study · the weights.
• The Narrator, a room for the spoken word: Fish Audio's Drama 3 through audio.cpp, on your card or the processor. Stresses from your marks, your dictionary and, for Ukrainian, lang-uk's index; Whisper hears every piece and reads again a word lost or added; a take fixed piece by piece; a book read part by part with [split]. Built with Fish Audio.
• Eight rooms in a new order, keys 1–8: the Musician, the Narrator, the Writer, the Refiner, the Artist, the Videographer, the Trainer and the Librarian. The Studio also speaks Hebrew, right to left.
• Two transcribers, Heretical (the default) and Refined, measured on our test songs before the other setups were left out.
• The models in one tree and a triple guard on every download: the pinned commit, the pinned size and the SHA-256 from the Studio's own manifest.
• The AppImage opens a window of its own first, offline too, and carries its own C library: measured on Debian 10 and CentOS 7.

Next, in 2.0.8: Higgs Audio v3 (Boson AI) as the Narrator's second engine, its weights fetched by your own hand from Boson's source; the Videographer's own tracks with templates for titles and karaoke.

To update: git pull or Pinokio's Update, or Engine → Updates; the AppImage is on the release page. Every change, line by line: CHANGELOG.

Two mini-guides came with it: Switching decoders and Reading a book with the Narrator. Questions and findings are welcome right here.