sh/chattts.py

import sys
import torch
import torchaudio

# --- 1. 将项目路径添加到Python的搜索路径中 ---
project_path = '/data/liyangzhuo/chattts/Awesome-ChatTTS-2-main'
if project_path not in sys.path:
    sys.path.insert(0, project_path)

import ChatTTS

# --- PyTorch 配置 ---
torch._dynamo.config.cache_size_limit = 64
torch._dynamo.config.suppress_errors = True
torch.set_float32_matmul_precision('high')

# --- 初始化并加载模型 ---
chat = ChatTTS.Chat()
# 确保这个路径是正确的，根据您之前的代码，这里应该是 /data/liyangzhuo/chattts/Awesome-ChatTTS-2-main/ChatTTS
# 我暂时先使用之前的路径，如果不对请修改
model_path = '/data/liyangzhuo/chattts/ChatTTS'
chat.load_models(source='local', local_path=model_path, compile=False)

# --- 定义文本输入 ---
texts = [
    "今天天气真不错，阳光暖洋洋的，感觉整个人都充满了活力。",
]

# --- 执行推理 ---
print("正在生成音频...")
wavs = chat.infer(texts, use_decoder=True) 
print("音频生成完毕。")

# --- 保存生成的第一个音频 ---
# 核心修正：直接转换，不需要 unsqueeze，因为输入已经是 2D 的了
output_path = "output_from_local_repo1.wav"
tensor_to_save = torch.from_numpy(wavs[0])

# 打印一下形状来确认
print(f"原始numpy数组形状: {wavs[0].shape}")
print(f"用于保存的Tensor形状: {tensor_to_save.shape}") # 现在应该输出 torch.Size([1, N])

torchaudio.save(output_path, tensor_to_save, 24000)
print(f"音频文件已成功保存至: {output_path}")