vocaboot — Third-party model attribution notice
================================================

This wav was produced using third-party model weights. The vocaboot code
itself is Apache-2.0, but the weights below are NOT — see each section.

----------------------------------------------------------------------
NSF-HiFiGAN ONNX vocoder (CC-BY-NC-SA-4.0)
----------------------------------------------------------------------

Component    : NSF-HiFiGAN vocoder, 44.1 kHz / 128-bin mel, F0-conditioned
Filename     : nsf_hifigan.onnx
Registry pin : openvpi-nsf-hifigan-onnx-20221211
SHA-256      : a3e26672a8c655e3faf65f31cb4339a7fbca7758ba86be9af89e03dced7c3fa4
Provider     : OpenVPI Community / DiffSinger Community
Artifact URI : https://github.com/openvpi/vocoders/releases/download/nsf-hifigan-v1/nsf_hifigan_onnx_20221211.zip
Source page  : https://github.com/openvpi/vocoders/releases
Project page : https://openvpi.github.io/vocoders/
License      : CC-BY-NC-SA-4.0
               https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode
Modifications: none — the file is loaded as-is, byte-identical to the
               upstream artifact (vocaboot.svs_engine._load_vocoder
               re-verifies the SHA-256 above on every load and refuses
               to run on mismatch). Per CC-BY-NC-SA-4.0 §3(a)(1)(B),
               this notice satisfies the "indicate if You modified the
               Licensed Material" clause.

Distribution clauses summarized (the legalcode is authoritative):

  1. Attribution — you must give appropriate credit, provide a link to
     the license, and indicate if changes were made. The credit line
     for this vocoder is:
         "NSF-HiFiGAN vocoder provided by the OpenVPI Community /
          DiffSinger Community, licensed under CC-BY-NC-SA-4.0."
  2. NonCommercial — you may not use the material for commercial
     advantage or monetary compensation. Personal smoke runs, research,
     and noncommercial OSS development are in scope.
  3. ShareAlike — if you remix, transform, or build upon the material
     (e.g. fine-tune the vocoder), the contributions must be distributed
     under the same license (CC-BY-NC-SA-4.0).
  4. No additional restrictions — you may not apply legal terms or
     technological measures that legally restrict others from doing
     anything the license permits.

Output wav status (vocaboot interpretation, conservative):

  - Creative Commons' public FAQ states that CC license conditions on
    the *model* do NOT automatically extend to model *outputs*
    (https://creativecommons.org/faq/, "Using CC-licensed Works for AI
    Training", May 2025). Under this reading, the resulting .wav is
    the user's copyright and is NOT automatically CC-BY-NC-SA-4.0.
  - HOWEVER, the act of *running* the NC weights is bound by the NC
    clause. Generating wavs for commercial use therefore requires the
    self-train path (see docs/ckpt_registry.md "Self-train path"); the
    --accept-nc opt-in is for personal, non-commercial use only.
  - Fine-tuning the vocoder or any CC-BY-NC-SA-4.0 acoustic produces a
    derivative work that DOES inherit CC-BY-NC-SA-4.0 by the
    ShareAlike clause.

----------------------------------------------------------------------
Reproducibility
----------------------------------------------------------------------

vocaboot enforces SHA-256 verification of the vocoder file on every
load (vocaboot.svs_engine._load_vocoder). The pin above is the
authoritative reference; a mismatch raises VocoderHashMismatchError and
refuses to synthesize.

This notice is emitted to stdout and persisted next to the output wav
whenever --accept-nc is passed. Removing or altering this notice does
not waive the CC-BY-NC-SA-4.0 obligations above.
