Spaces:
Build error
Build error
Upload app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,132 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import gradio as gr
|
| 3 |
+
from gradio_client import Client, handle_file
|
| 4 |
+
import shutil
|
| 5 |
+
from groq import Groq
|
| 6 |
+
|
| 7 |
+
groq_client = Groq()
|
| 8 |
+
|
| 9 |
+
def run_tts(ref_audio_file, ref_text, gen_text):
|
| 10 |
+
"""
|
| 11 |
+
调用 TTS 模型并保存结果
|
| 12 |
+
"""
|
| 13 |
+
try:
|
| 14 |
+
if ref_audio_file is None:
|
| 15 |
+
return "请上传参考音频文件。", None # 返回错误消息和None
|
| 16 |
+
|
| 17 |
+
ref_audio_path = ref_audio_file.name # 获取上传文件的路径
|
| 18 |
+
|
| 19 |
+
# 初始化客户端,不使用 token
|
| 20 |
+
client = Client("abidlabs/E2-F5-TTS")
|
| 21 |
+
|
| 22 |
+
# 调用 /infer 端点
|
| 23 |
+
print("调用 /infer 端点...")
|
| 24 |
+
result = client.predict(
|
| 25 |
+
ref_audio_orig=handle_file(ref_audio_path), # 使用上传的音频文件
|
| 26 |
+
ref_text=ref_text,
|
| 27 |
+
gen_text=gen_text,
|
| 28 |
+
exp_name="F5-TTS",
|
| 29 |
+
remove_silence=False,
|
| 30 |
+
cross_fade_duration=0.15,
|
| 31 |
+
api_name="/infer"
|
| 32 |
+
)
|
| 33 |
+
print("infer端点返回结果:", result)
|
| 34 |
+
|
| 35 |
+
# 处理返回结果,将文件保存到当前目录
|
| 36 |
+
output_file = None
|
| 37 |
+
if isinstance(result, tuple):
|
| 38 |
+
for item in result:
|
| 39 |
+
if isinstance(item, str) and item.lower().endswith(".wav"):
|
| 40 |
+
if os.path.exists(item): # 确保item是一个文件路径且存在
|
| 41 |
+
# 获取文件名
|
| 42 |
+
filename = os.path.basename(item)
|
| 43 |
+
# 构造新的保存路径,保持在当前目录
|
| 44 |
+
new_path = os.path.join(".", filename)
|
| 45 |
+
# 复制文件到新路径
|
| 46 |
+
shutil.copy2(item, new_path)
|
| 47 |
+
print(f"音频文件 '{filename}' 已保存到: {new_path}")
|
| 48 |
+
output_file = new_path
|
| 49 |
+
break #只保留一个音频文件路径
|
| 50 |
+
else:
|
| 51 |
+
print(f"文件路径不存在,跳过: {item}")
|
| 52 |
+
elif isinstance(item,str):
|
| 53 |
+
print(f"跳过非音频文件: {item}")
|
| 54 |
+
elif isinstance(result, str) and result.lower().endswith(".wav"):
|
| 55 |
+
if os.path.exists(result):
|
| 56 |
+
# 获取文件名
|
| 57 |
+
filename = os.path.basename(result)
|
| 58 |
+
# 构造新的保存路径,保持在当前目录
|
| 59 |
+
new_path = os.path.join(".", filename)
|
| 60 |
+
# 复制文件到新路径
|
| 61 |
+
shutil.copy2(result, new_path)
|
| 62 |
+
print(f"音频文件 '{filename}' 已保存到: {new_path}")
|
| 63 |
+
output_file = new_path
|
| 64 |
+
else:
|
| 65 |
+
print(f"文件路径不存在,跳过: {result}")
|
| 66 |
+
elif isinstance(result,str):
|
| 67 |
+
print(f"跳过非音频文件: {result}")
|
| 68 |
+
else:
|
| 69 |
+
print(f"跳过非字符串/元组类型的返回值: {result}")
|
| 70 |
+
|
| 71 |
+
if output_file:
|
| 72 |
+
return output_file , ref_text # 返回音频文件路径字符串和修改后的参考文本
|
| 73 |
+
else:
|
| 74 |
+
return "未生成音频文件。", ref_text # 返回错误提示和修改后的参考文本
|
| 75 |
+
except FileNotFoundError as e:
|
| 76 |
+
return f"发生错误:{e}", ref_text
|
| 77 |
+
except Exception as e:
|
| 78 |
+
return f"发生未知错误:{e}", ref_text
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
def transcribe_audio(audio_file):
|
| 82 |
+
"""
|
| 83 |
+
使用 Groq 进行语音识别
|
| 84 |
+
"""
|
| 85 |
+
try:
|
| 86 |
+
if audio_file is None:
|
| 87 |
+
return "请上传参考音频文件。", None
|
| 88 |
+
|
| 89 |
+
audio_path = audio_file.name
|
| 90 |
+
with open(audio_path, "rb") as file:
|
| 91 |
+
transcription = groq_client.audio.transcriptions.create(
|
| 92 |
+
file=(audio_path, file.read()),
|
| 93 |
+
model="whisper-large-v3-turbo",
|
| 94 |
+
language="zh",
|
| 95 |
+
)
|
| 96 |
+
return transcription.text, audio_file # 返回识别文本和音频文件
|
| 97 |
+
except Exception as e:
|
| 98 |
+
return f"语音识别失败: {e}", None
|
| 99 |
+
|
| 100 |
+
|
| 101 |
+
def update_ref_text(audio_file, ref_text_box):
|
| 102 |
+
"""
|
| 103 |
+
语音识别并更新参考文本
|
| 104 |
+
"""
|
| 105 |
+
transcribed_text, audio_file = transcribe_audio(audio_file)
|
| 106 |
+
return transcribed_text, audio_file
|
| 107 |
+
|
| 108 |
+
|
| 109 |
+
with gr.Blocks(title="快速语音合成") as iface:
|
| 110 |
+
gr.Markdown("上传参考语音和输入参考及生成文本,生成相应的语音。")
|
| 111 |
+
ref_audio_input = gr.File(file_types=["audio"], label="参考语音 (上传音频自动识别)")
|
| 112 |
+
ref_text_input = gr.Textbox(label="参考语言 (文本)")
|
| 113 |
+
gen_text_input = gr.Textbox(label="生成语言 (文本)")
|
| 114 |
+
|
| 115 |
+
audio_output = gr.Audio(label="生成的语音 (下载)")
|
| 116 |
+
|
| 117 |
+
ref_audio_input.upload(
|
| 118 |
+
update_ref_text,
|
| 119 |
+
inputs=[ref_audio_input, ref_text_input],
|
| 120 |
+
outputs=[ref_text_input, ref_audio_input]
|
| 121 |
+
)
|
| 122 |
+
|
| 123 |
+
|
| 124 |
+
btn = gr.Button("合成")
|
| 125 |
+
btn.click(
|
| 126 |
+
run_tts,
|
| 127 |
+
inputs=[ref_audio_input, ref_text_input, gen_text_input],
|
| 128 |
+
outputs=[audio_output,ref_text_input],
|
| 129 |
+
)
|
| 130 |
+
|
| 131 |
+
if __name__ == "__main__":
|
| 132 |
+
iface.launch(share=True)
|