The Voice License

v0.1 draft · applies to /api/v1/voice · not legal advice — a formal agreement is available on request

The PhotoPrompting voice is distilled from a living corpus of real one-line takes — tweetable comments on each element of a spotify song — written by this community and re-distilled weekly. A style can’t be copyrighted. A corpus, a distilled profile, and metered access can be owned and licensed. This page is the deal.

what your key grants

• Outputs are yours. Text generated through your key may be used commercially — product copy, flavor text, posts, print — with no attribution required (though “voice: hermit / photoprompting.org” is appreciated).
• Provenance included. Every response is stamped with the voice version and corpus size, so you can prove where your words came from.
• Metered, not rented. Quotas are per key per day. What you generate within quota is yours forever — the license on an output survives key expiry.

what stays ours

• The corpus. The takes themselves are copyrighted works of their authors. A key is not a license to scrape, redistribute, or train on them.
• The profile. The distilled voice document is our copyrighted expression. It is never included in responses and may not be extracted, reconstructed, or redistributed.
• The voice itself. You may not use outputs to train, fine-tune, or prompt-engineer a model or product whose purpose is to reproduce this voice without a key. This is the anti-cloning clause — contract does here what copyright cannot.
• The name. “Hermit” and “PhotoPrompting” identify this voice and service; don’t imply endorsement.

what is spotify’s

• Cover art, metadata, previews. Every cover, title, artist, and 30-second clip on this site is Spotify’s content, shown with the Spotify mark and a link back to the track, as Spotify’s Developer Policy asks. If you show a cover or a clip you got from this API, keep that link back with it, and never offer the art or the clip on its own.
• The 30 seconds. They play in Spotify’s own embedded player, from Spotify’s servers, never copied. A stream from any other service is never placed beside Spotify’s cover art or song details.
• Never trained on. The voice learns from the wall’s human lines and nothing of Spotify’s — no audio, no previews, no lyrics. Title and artist are the only Spotify facts that reach it.

the boring but honest part

• Don’t generate content that impersonates real people, or use the voice for spam, harassment, or deception about authorship where disclosure matters.
• Keys are revocable for breach; generated-before-revocation outputs keep their license.
• The voice evolves with the corpus. Pin nothing to today’s register; that’s the point of a living voice.

getting a key

Request a key here — reviewed personally, issued by email. Try the voice first, free: the doors and docs cover the unmetered surfaces.

◈ endpoint: POST /api/v1/voice · body: { prompt, form?, length?, best? } · auth: x-api-key