metadata
license: mit
language:
- en
tags:
- accent-conversion
- speech
- conformer
- flow-matching
library_name: cosyaccent
CosyAccent
Model weights for CosyAccent: Duration-Controllable Accent Normalization Using Source-Synthesis Training Data (ICASSP 2026).
- Paper: https://arxiv.org/abs/2602.19166v1
- Data: https://huggingface.co/datasets/Piping/L2-LibriTTSR
- Code & documentation: https://github.com/P1ping/CosyAccent
Usage
See the GitHub repository for full installation and inference instructions.
Quick start:
pip install -r requirements.txt
python infer_wav.py \
--source_wav /path/to/source.wav \
--output_wav outputs/result.wav
Citation
@inproceedings{bai2026cosyaccent,
title={CosyAccent: Duration-Controllable Accent Normalization Using Source-Synthesis Training Data},
author={Bai, Qibing and Shi, Shuhao and Wang, Shuai and Ju, Yukai and Wang, Yannan and Li, Haizhou},
booktitle={ICASSP 2026},
year={2026}
}