Licenses and rights

Licenses

The dataset license is decided. The model license is not. Voice and publicity rights are handled in the consent form, separately from copyright.

Dataset: Creative Commons Attribution 4.0

Decided November 15, 2026. The recordings, transcripts, and metadata are planned for release under CC BY 4.0.

Anyone may copy, adapt, and reuse the dataset for any purpose, including commercially, if they give attribution. The project asks that credit go in the dataset or model card and in docs — not inside each generated audio clip. That is treated as reasonable attribution under CC BY 4.0 section 3(a)(2).

Credit line (use this exactly, once the contributors URL exists):

OpenSanta Dataset by OpenSanta contributors (contributors URL coming soon), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

The full explanation, including the dataset notice draft, is in the dataset license notes.

Model license: not yet decided

The trained voice will be a derivative of whatever base TTS model we fine-tune, so that base license limits what we can choose. The current recommendation is an OpenRAIL-M style license — free for commercial use, with rules against impersonation and fraud that travel with the model — and Apache-2.0 plus a published acceptable-use request as the fallback.

This is still open. See the comparison and draft notices in the model license notes.

Voice and publicity rights

CC BY 4.0 covers copyright only. It does not license personality or publicity rights (section 2(b)(1)). The consent form grants voice and publicity permissions separately, so the recordings and the model can be used as the project describes.

The script

All 280 lines in the recording script were written fresh for this project and will be released with the dataset under CC BY 4.0. See the recording script notes.