🎤 RIFT-SVC 歌声音色转换 （七海Nanami demo）

使用 RIFT-SVC 模型将歌声或语音转换为七海Nanami的音色

🔗 想要微调自己的说话人？ 请访问 RIFT-SVC GitHub 仓库 获取完整的训练和微调指南。

🎤 数据来源说明： 该demo数据来源为b站上快速爬取的约30分钟七海唱歌片段，直接分离人声后进行训练，没有额外筛选。

📝 注意： 为获得最佳效果，请使用背景噪音较少的干净音频。最大音频长度为5分钟。建议用较短的音频测试避免平台意外中断任务。

📥 输入
目标说话人
-12 12
8 64
音高滤波

0=无，1=轻度过滤，2=强力过滤（有助于解决断音/破音问题）
1 64
0 1
0 2
0 1
0 1
用于反向引导的内容向量下采样率

更高的值（可能）可以提高内容清晰度。
-60 -20
1000 10000
10 500
1 20
10 1000

📤 输出

❌ 加载模型出错: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

详细信息: Traceback (most recent call last): File "/home/user/app/app.py", line 82, in initialize_models svc_model, vocoder, rmvpe, hubert, rms_extractor, spk2idx, dataset_cfg = load_models(model_path, device, use_fp16) File "/home/user/app/infer.py", line 56, in load_models rmvpe = RMVPE(model_path="pretrained/rmvpe/model.pt", hop_length=160, device=device) File "/home/user/app/rift_svc/rmvpe/inference.py", line 14, in init ckpt = torch.load(model_path, weights_only=True) File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 1462, in load return _load( File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 1964, in _load result = unpickler.load() File "/usr/local/lib/python3.10/site-packages/torch/_weights_only_unpickler.py", line 512, in load self.append(self.persistent_load(pid)) File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 1928, in persistent_load typed_storage = load_tensor( File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 1900, in load_tensor wrap_storage=restore_location(storage, location), File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 693, in default_restore_location result = fn(storage, location) File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 631, in _deserialize device = _validate_device(location, backend_name) File "/usr/local/lib/python3.10/site-packages/torch/serialization.py", line 600, in _validate_device raise RuntimeError( RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

🔍 快速提示

  • 音调调整： 以半音为单位上调或下调音高。
  • 推理步骤： 步骤越多 = 质量越好但速度越慢。
  • 音高滤波： 有助于提高具有挑战性的音频中的音高稳定性。
  • 批处理大小： 值越大 = 转换越快，但需要更多GPU内存。遇到内存不足时降低此值。
  • CFG参数： 调整转换质量和音色。