Cadent — third-party model notices
=====================================

Cadent ships no model weights. It downloads the model you pick, once, from
Hugging Face, and the licences below are the ones that then apply to what is on
your disk. They are recorded here because two of them ask to be.

Speech models
-------------

Whisper (`tiny.en`, `base.en`, `small.en`, `medium.en`, `large-v3` and the
`distil-` distillations) — MIT. Original model by OpenAI; the CTranslate2
conversions Cadent downloads are published by Systran, and the distilled
variants by Distil-Whisper.

NVIDIA Parakeet TDT 0.6B v2 / v3 — CC-BY-4.0. Attribution required:
these models are by NVIDIA (https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2,
https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3).

Cleanup models
--------------

**Built with Llama.** The "Fastest" and "Balanced" cleanup rungs are
Llama 3.2 1B Instruct and Llama 3.2 3B Instruct, built with Meta Llama 3.2 and
used under the Llama 3.2 Community License Agreement
(https://www.llama.com/llama3_2/license/). Llama 3.2 is licensed under the
Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights
Reserved. The GGUF quantisations Cadent downloads are published by
bartowski.

Qwen3 1.7B and Qwen3 4B Instruct 2507 — Apache-2.0. Copyright © Alibaba Cloud.
The GGUF quantisations Cadent downloads are published by unsloth.

Choosing your own GGUF file puts its licence entirely in your hands; Cadent
neither fetches nor inspects it.
