Werea-TSS
Copyright 2026 Werea

This product includes modified weights derived from FreyaTTS-small:
Copyright 2026 Freya Voice AI, licensed under the Apache License 2.0.
Source: https://github.com/freyavoiceai/FreyaTTS
Base model: https://huggingface.co/freyavoice/Freya-TTS
Pinned source commit: 146d36c1cb6660646be57d31339db4eed9315de3

Modifications by Werea: conservative full-parameter fine-tuning on a
Werea-authored Turkish text corpus with synthetic teacher audio, checkpoint
selection on an isolated Turkish evaluation set, and model-card additions.

The frozen audio decoder is AudioVAE2 from openbmb/VoxCPM2, also distributed
under Apache License 2.0: https://huggingface.co/openbmb/VoxCPM2