---
license: mit
---


# pitch_extractor.safetensors

Pretrained pitch extractor model for use when making a new base model with StyleTTS 2 or Stylish TTS

- Taken from the StyleTTS 2 repo: https://github.com/yl4579/StyleTTS2
- This is a Joint Detection and Classification network for melody
- Original Paper: https://www.mdpi.com/2076-3417/9/7/1324
- Original Code: https://github.com/keums/melodyExtraction_JDC
- Trained on LibriTTS corpus

# rmvpe.safetensors

Pretrained RMVPE model for use in training other models.

- Taken from the Applio repository: https://huggingface.co/IAHispano/Applio/blob/main/Resources/rmvpe.pt
- Based on "RMVPE: A Robust Model for Vocal Pitch Estimation in Polyphonic Music" by Haojie Wei, Xueke Cao, Tangpeng Dan, Yueguo Chen [Paper](https://www.isca-archive.org/interspeech_2023/wei23b_interspeech.html)