SecureLinkAI voice engine pack

These files are the optional High quality voice engine of SecureLinkAI. They are downloaded automatically by the SecureLinkAI installer. You do not need to download them by hand.

Licences of the parts

Component	Licence
Seed-VC (code and model weights)	GPL-3.0, https://github.com/Plachtaa/seed-vc
XLS-R content encoder (facebook/wav2vec2-xls-r-300m)	Apache-2.0
CAM++ speaker model (funasr/campplus)	Apache-2.0
HiFT vocoder (FunAudioLLM/CosyVoice-300M)	Apache-2.0
PyTorch, Transformers and other Python libraries	BSD / Apache-2.0 / MIT (full list in NOTICE.txt inside the pack)
Reference voices	CC BY 4.0, CSTR VCTK Corpus, University of Edinburgh

Source code

The GPL-licensed engine program is shipped as readable Python source inside the pack. A written source offer is in SOURCE-OFFER.txt inside the pack, and the upstream project is at https://github.com/Plachtaa/seed-vc

Contents

The pack is split into parts of roughly 1 GB (engine-001.zip, runtime-001.zip, models-001.zip, torch-001.zip, ...) and a signed list called manifest.json. The SecureLinkAI installer reads manifest.json, downloads each part (resuming if the connection drops), checks it, and unpacks it.