---
language: 
  - zh
  - en
tags:
  - speech-synthesis
  - speech-to-speech
  - voice-conversion
  - pytorch
  - audio
  - chinese-tts
  - multi-speaker
  - convolution
  - encoder-decoder
license: apache-2.0
datasets:
- vctk
library_name: pytorch
---

# Convbased

本项目专注于训练高质量的预训练底模，为语音转换任务提供强大的基础模型支持。

| 特征提取 | 声码器 | 40k | 48k |
|-----------|--------|-----|-----|
| contentvec | hifigannsf | ❌ | ✅ |
| contentvec | sifigan | ❌ | ✅ |
| contentvec | refinegan | ❌ | ❌ |
| contentvec | hifiganmrf | ❌ | ❌ |
| spin | hifigannsf | ❌ | ❌ |
| spin | sifigan | ❌ | ✅ |
| spin | refinegan | ❌ | ❌ |
| spin | hifiganmrf | ❌ | ❌ |
| chinese-hubert-base | hifigannsf | ❌ | ✅ |
| chinese-hubert-base | sifigan | ❌ | ❌ |
| chinese-hubert-base | refinegan | ❌ | ❌ |
| chinese-hubert-base | hifiganmrf | ❌ | ❌ |


*致力于推进中文语音合成技术的发展，该底模已用于微调大部分模型于 [Convbased Studio](https://weights.chat/)*