ONNX Model Zoo

happyme531/MeloTTS-RKNN2

By happyme531 · ↓ 0 · ♥ 0 · Updated 2025-07-01

What's in this repo

This page lists the files in happyme531/MeloTTS-RKNN2, including decoder.onnx, encoder.onnx.

  • 2 ONNX models (.onnx files)
  • 73 files, including 42 .py, 9 .txt, 8 .pyc, 4 .json, 2 .onnx, 1 .md
  • Last updated 2025-07-01

ONNX files

File Size Download
decoder.onnx — Download
encoder.onnx — Download

How to use

These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.

README

MeloTTS-RKNN2

(English README see below)

在RK3588上运行MeloTTS文字转语音模型!

  • 推理速度(RK3588): 约5倍速
  • 内存占用(RK3588): 约0.2GB

使用方法

  1. 克隆或者下载此仓库到瑞芯微SoC的系统上.

  2. 安装依赖

pip install -r requirements.txt
pip install rknn-toolkit-lite2
  1. 运行
python melotts_rknn.py -s "你想要生成的文本"

模型转换

  1. 安装依赖
pip install -r requirements.txt
pip install rknn-toolkit2==2.3.0
  1. 转换模型
python convert_rknn.py

已知问题

  • 和原项目一样，Encoder部分并没有使用NPU加速，但是耗时不大，应该不会对推理速度有太大影响。

参考

English README

MeloTTS-RKNN2

Run the MeloTTS text-to-speech model on RK3588!

  • Inference speed (RK3588): about 5x real-time
  • Memory usage (RK3588): about 0.2GB

Usage

  1. Clone or download this repository to your Rockchip SoC system.

  2. Install dependencies

pip install -r requirements.txt
pip install rknn-toolkit-lite2
  1. Run
python melotts_rknn.py -s "The text you want to generate."

Model Conversion

  1. Install dependencies
pip install -r requirements.txt
pip install rknn-toolkit2==2.3.0
  1. Convert the model
python convert_rknn.py

Known Issues

  • Same as the original project, the Encoder part is not accelerated by the NPU. However, its processing time is short and should not significantly affect the inference speed.

References

Model info

onnx rknn license:agpl-3.0 region:us
↓ 0 ♥ 0

Created: 2025-05-07

Updated: 2025-07-01

View on Hugging Face

Files (73)

.gitattributes
README.md
convert_rknn.py
decoder.onnx ONNX
decoder.rknn
encoder.onnx ONNX
english_utils/__init__.py
english_utils/__pycache__/__init__.cpython-39.pyc
english_utils/__pycache__/abbreviations.cpython-39.pyc
english_utils/__pycache__/number_norm.cpython-39.pyc
english_utils/__pycache__/time_norm.cpython-39.pyc
english_utils/abbreviations.py
english_utils/number_norm.py
english_utils/time_norm.py
g.bin
lexicon.txt
melotts_rknn.py
output.wav
requirements.txt
text/__init__.py
text/__pycache__/__init__.cpython-39.pyc
text/__pycache__/chinese_bert.cpython-39.pyc
text/__pycache__/cleaner.cpython-39.pyc
text/__pycache__/symbols.cpython-39.pyc
text/chinese.py
text/chinese_bert.py
text/chinese_mix.py
text/cleaner.py
text/cleaner_multiling.py
text/cmudict.rep
text/cmudict_cache.pickle
text/english.py
text/english_bert.py
text/english_utils/__init__.py
text/english_utils/abbreviations.py
text/english_utils/number_norm.py
text/english_utils/time_norm.py
text/es_phonemizer/__init__.py
text/es_phonemizer/base.py
text/es_phonemizer/cleaner.py
text/es_phonemizer/es_symbols.json
text/es_phonemizer/es_symbols.txt
text/es_phonemizer/es_symbols_v2.json
text/es_phonemizer/es_to_ipa.py
text/es_phonemizer/example_ipa.txt
text/es_phonemizer/gruut_wrapper.py
text/es_phonemizer/punctuation.py
text/es_phonemizer/spanish_symbols.txt
text/es_phonemizer/test.ipynb
text/fr_phonemizer/__init__.py
text/fr_phonemizer/base.py
text/fr_phonemizer/cleaner.py
text/fr_phonemizer/en_symbols.json
text/fr_phonemizer/example_ipa.txt
text/fr_phonemizer/fr_symbols.json
text/fr_phonemizer/fr_to_ipa.py
text/fr_phonemizer/french_abbreviations.py
text/fr_phonemizer/french_symbols.txt
text/fr_phonemizer/gruut_wrapper.py
text/fr_phonemizer/punctuation.py
text/french.py
text/french_bert.py
text/japanese.py
text/japanese_bert.py
text/ko_dictionary.py
text/korean.py
text/opencpop-strict.txt
text/spanish.py
text/spanish_bert.py
text/symbols.py
text/tone_sandhi.py
tokens.txt
utils.py

Related models

More from happyme531

More models

No task type declared.

Popular models