The idea
Take one singer’s recorded corpus, about four hours of singing, and grow a whole choir out of it. Not by imitating anyone else’s voice, but by displacing the register of the one voice until distinct characters emerge, then giving each of them a role.
The cast
| Singer | Job |
|---|---|
| The Abyss | true bass, the substructure |
| The Baritone | the core male anchor |
| The Tenor | cut, height, urgency |
| The Wraith | air and ghost layer |
| The Heretic | force, grit, attack |
| The Monk | purity and stability |
| The Young Voice | the light upper voice |
| The Other | the nonhuman specialist |
Voicings are written as major, minor, power, and octaves. Modes are human, cinematic, and the original undead register. A section can be scored without a given singer, and low-choir subsets are a first-class arrangement.
The production line
The choir sits inside a wider FL Studio system: a music architect for structure, a beat builder, a score builder, an arrangement editor, and a semantic sound-design engine over DUNE 3’s 1,762 indexed patches. Renders run from the command line in seconds, which makes the listen-fix-render loop fast enough to actually use.
What is real and what is not
- Real: two finished vocal-trance tracks, Aurora Velocity and Midnight Signal, approved as pleasant to the ear at commercial quality by the only judge that counts. The corpus pipeline. The base voice model. The undead-register choir.
- Not yet: the second generation of singer identities. Eight singers were trained on speech for a singing project, and they failed the blind identity test. They were retired as evidence, not shipped. The rebuild uses the singing corpus and a different identity method.
Audio will be posted here when there is something to hear that survives a second listen.
Field reports