dsh-sovits-tts — third-party notices
====================================

This repository is a derivative work: the inference code under `engine/vendor/`
is copied (with deletions and small fixes) from the upstream project listed
below. The remaining code (the DSH bundle in the repository root and the
pipeline in `engine/engine.py`, `engine/worker.py`) is original work.

No model weights are distributed with this repository
(`G_*.pth`, `hubert-soft-*.pt`, `config.json` belong to whoever trained them).


1. so-vits-svc (32k branch) — MIT
---------------------------------
Files copied from upstream into `engine/vendor/`:

  engine/vendor/models.py
  engine/vendor/modules.py
  engine/vendor/attentions.py
  engine/vendor/commons.py
  engine/vendor/hubert/hubert_model.py
  engine/vendor/vdecoder/hifigan/models.py
  engine/vendor/vdecoder/hifigan/env.py

Files rewritten or modified after the upstream original:

  engine/vendor/utils.py                      reduced to f0_to_coarse /
                                              load_checkpoint / get_hparams_from_file
  engine/vendor/vdecoder/hifigan/utils.py     matplotlib import moved into the
                                              plotting function; np.int -> np.int64

Upstream project: SoftVC VITS Singing Voice Conversion
  https://github.com/innnky/so-vits-svc  (32k branch)
  Copyright (c) 2021 Jingyi Li — MIT License (full text in LICENSE).


2. NSF HiFiGAN vocoder — upstream of (1)
----------------------------------------
`engine/vendor/vdecoder/hifigan/` is the NSF HiFiGAN decoder that so-vits-svc
bundles. Its original upstream is:
  https://github.com/openvpi/DiffSinger  (modules/nsf_hifigan)
It reached this repository through (1), under (1)'s MIT license file.


3. HuBERT-Soft content encoder — MIT, bshall/hubert
---------------------------------------------------
`engine/vendor/hubert/hubert_model.py` is the pure-PyTorch HuBERT-Soft model
definition bundled by (1). The checkpoint `hubert-soft-0d54a1f4.pt` is NOT
included; download it from:
  https://github.com/bshall/hubert/releases/tag/v0.1
  https://github.com/bshall/hubert


4. Source speech
----------------
  * edge-tts (Microsoft Edge online TTS) — https://github.com/rany2/edge-tts
  * Windows SAPI (System.Speech / SAPI.SpVoice) via comtypes, local only.


5. Voice models and character voices
------------------------------------
The voices this plugin can re-voice into are so-vits-svc checkpoints trained by
their owners. Their training data, character likeness and usage rights are NOT
covered by this repository's license. Obtain permission from the rights holders
before synthesizing with someone else's voice model in public.

If you are an upstream author and believe a file here is used inappropriately,
open an issue and the offending files will be removed or the repository made
private.