Repository navigation
Ruach Studio 2.0.0-rc7
Pre-release
Pre-release
The seventh release candidate: songs edited as the model writes them, compare with a take's namesakes, the studio's own LoRA Trainer made whole, covers by the cards on Qwen-Image 2512 (HERESY 1392–1425).
The Creator
- Replace a section: a right click on a section of the player's wave, its words changed as you will, sung anew between the rests the ear finds; the rest of the take is kept to the sample, so the seam goes unheard. Several sections in one go; the versions named by section and stretch (Chorus 1:02–1:25 anew A) and in Takes as they land.
- Two guidances: the style's pull and the lyrics' hold, each its own scale, in tenths.
- Free tempo: the score's tempo out before the music, on by the Style's own words.
- Compare: a take with its namesakes, or two to four checked, in one player from the same second: 1200 px, the playing version's wave as the seek bar, thumbs, star and «Hide disliked».
The LoRA Trainer
- Epochs instead of steps; score mode (Full, Melody: each song's sheet transcribed once and learned with it); the separation of vocals and music (BS-Roformer once a set, three terms in the loss). AI-Toolkit is no longer needed by the studio.
The Artist and the Refiner
- Covers by the cards: Qwen-Image 2512 where a card holds it whole (Q6 from 20 GB, Q4 from 16 GB; its text encoder leaves the card before the painter comes), else Krea 2 Muse. A cover copied to a take's namesakes from its menu.
- Colour in the Refiner: vinyl (surface, fine crackle, a click now and then, the player's rumble), tape or room, a layer under the song before the loudness.
- The Artist in Depth in all twelve languages.
- Engine: Compute has the VAE beside the model; a card's VRAM limit sets the memory knobs (the preset is gone); the machine scan suggests Qwen-Image 2512 where it fits.
Install / update
- New: INSTALL.md, or Pinokio on Linux: download it from
https://github.com/igrbible/Ruach_Studioand press Install. - From rc6: Engine → Updates in the page, or Pinokio's Update. The Qwen painters come from Engine → Models and LoRAs.
Linux with an NVIDIA card (12 GB and more). Windows 11 through WSL2: INSTALL_WINDOWS.md.