Skip to content
Arkancia

Manifesto

The era of hiding the AI is over.

Arkancia is an independent studio building an AI-native label the way labels should be built — disclosed, owned, fairly credited, and ethical by design.

A cinematic archway opening onto a violet-lit horizon — the Arkancia thesis.

The music industry has been quietly adopting AI for two decades — auto-tune, stem separation, mastering, now full-stack generation. The question is no longer whether AI touches the product, but whether anyone admits it.

Arkancia's position: AI is a writing surface, the artist is the edit, and the label is the container that enforces rights and credits. We disclose — on the page, in the metadata, in the paperwork.

Thesis

Three layers. One company.

1 · Owned IP

Fictional artists with real catalogues. Masters, publishing, and likeness owned by Arkancia. Single-point sync licensing, no rightsholder roulette.

2 · Blueprint

The exact operating manual — positioning, production, rights, distribution, sync, scaling. Book, cohort, done-with-you, enterprise.

3 · Video extension

Same thesis, short-form video IP. In pre-production. A cross-medium label where characters travel across music and screen.

Ethics

Our red lines

  • No voice cloning of identifiable real persons — living or dead. Ever.
  • Full AI disclosure on every artist page, every release, every metadata tag. Required by the EU AI Act; we treat it as a baseline, not a ceiling.
  • Human collaborators paid transparently — songwriters, producers, vocalists, visual artists. Credits published publicly.
  • Training data provenance: we only use models and datasets with documented, licensable provenance for commercial output.
Read the full AI disclosure

Founder

An independent studio

Arkancia is a founder-led independent studio. One person at the wheel, a small network of trusted collaborators on production, clearance, and design. No venture money. No shareholder pressure to cut corners on disclosure.

FAQ

Ethics & approach

How does Arkancia handle the EU AI Act and artist consent?
Every voice model we ship is trained only on recordings the performing artist has explicitly licensed to us, with written contracts and revenue share. We register each synthetic voice as an AI-generated work per the EU AI Act's transparency requirements, and we publish an ethics note on every artist page. If a source artist withdraws consent, we retire the model.
How are the AI voice models trained?
Each model is fine-tuned on between 20 and 80 hours of newly-commissioned vocal sessions, performed by named human singers under explicit license. We never train on publicly-scraped recordings, never clone identifiable living artists without written consent, and we keep the source stems and contracts on file for audit. Every release credits the source performers and shares revenue with them.
Do you split publishing with the source performers?
Yes — every track where an AI voice model is used carries a transparent split with the human performers who trained that model, registered with SACEM / PRS / ASCAP on our side, and paid out quarterly. The exact split depends on the artist project and is documented in each performer's licensing contract.