Geukrak/onnx_zundamon

what is this model for?

this model repository has been created for voiceger program which was created by zunzun99 and yh00722. Since this repository is outdated, I'm currently doing some improvements and optimization.

Model Information Summary

1. General Information

• Supported Languages: Japanese (ja), Chinese (zh)
• Library: Transformers
• Pipeline Tag: Text-to-Speech

2. Base Models & Sources The supplied models are sourced from the following repositories:

• Hubert Model: TencentGameMate/chinese-hubert-base
• VITS & GPT Models: zunzunpj/zundamon_GPT-SoVITS
• RoBERTa Model: hfl/chinese-roberta-wwm-ext-large

3. Optimization

• The GPT and VITS model has been exported to an ONNX format, allowing full onnx pipeline support for fastest experience.
• converted BERT and HUbert model into safetensors format to improve security and zero-copy loading.

4. License Information The licenses for these models are as follows:

• chinese-roberta: Apache-2.0 License
• chinese-hubert-base: MIT License
• GPT and VITS Finetuned Models: Custom License (Check the license details here: https://zunko.jp/con_ongen_kiyaku.html)

5. lisence for this model

these converted models are deployed with AGPL-3 open source lisence. you can freely use everything in this repositories within lisence agreements.

additional links and information.

(my forked repo) (original repo) (voiceger program official site)