Aalto-Speech-Synthesis/HiFi-Glot

HiFi-Glot: High-Fidelity Neural Formant Synthesis with Differentiable Resonant Filters

This is the official Hugging Face model repository for the paper "[HiFi-Glot: High-Fidelity Neural Formant Synthesis with Differentiable Resonant Filters](TBD)", which is the first end-to-end neural formant synthesis system that achieves high perceptual quality and precise formant control.

For more details, please visit our GitHub Repository.

How to use

You need to put the pretrained checkpoints in:

  HiFi-Glot/experiments

of our official repository, and then follow the instructions written in the repository to resume, finetune, and inference our pretrained checkpoints.

Citation

@article{afgen,
  title        = {HiFi-Glot: High-Fidelity Neural Formant Synthesis with Differentiable Resonant Filters},
  author       = {Yicheng Gu and Pablo Pérez Zarazaga and Chaoren Wang and Zhizheng Wu and Gustav Eje Henter and Zofia Malisz and Lauri Juvela},
  year         = {2025},
  journal      = {TBD},
}