app.py

import gradio as gr
import os
import consts
from utils_base import get_dataset_list, get_model_list
#from utils_label import auto_label, delete_dataset
#from utils_sambert import train, infer, delete_model

os.system("pip install --upgrade Cython==0.29.35")
os.system("pip install pysptk --no-build-isolation")
os.system("pip install kantts -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html")
os.system("pip install tts-autolabel -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html")

def auto_label(a, b):
    return "this is", "a test"

def delete_dataset():
    return "test"

def train(a, b, c):
    return "this is", "a", "test"

def infer(a, b):
    return "this is a test"

def delete_model():
    return "test"

def refresh():
  return gr.update(choices=get_dataset_list()), gr.update(choices=get_model_list())

# gradio server ---------------------------
with gr.Blocks() as server:
  # 面板说明
  gr.Markdown("# <center>🌊💕🎶 Sambert UI 声音克隆</center>")
  gr.Markdown("## <center>🌟 - 训练5分钟，通话不限时！AI真实拟声，支持中英双语！ </center>")      
  gr.Markdown("### <center>🍻 - 更多精彩应用，尽在[滔滔AI](http://www.talktalkai.com)；滔滔AI，为爱滔滔！💕</center>")

  # 标记
  gr.Markdown('## 数据标注')
  with gr.Row():
    label_audio_input = gr.Audio(type='filepath', label='请上传一段长音频（一分钟左右即可）')
    label_name_input = gr.Textbox(label='角色命名')
    label_status_output = gr.Textbox(label='标注状态')
    label_btn = gr.Button('开始标注', variant='primary')

  # 训练
  gr.Markdown('## 训练')
  with gr.Row():
    train_dataset_input = gr.Radio(label='角色选择', choices=get_dataset_list())
    train_name_input = label_name_input
    train_steps_input = gr.Number(label='训练步数, 需要为20的整数倍')
    train_status_output = gr.Text(label='训练状态')
    train_btn = gr.Button('开始训练')
    dataset_delete_btn = gr.Button('删除数据集', variant='stop')

  # 推理
  # 参考 https://mdnice.com/writing/a40f4bcd3b3e40d8931512186982b711
  # 使用 gr.update 实现对应的联动效果
  gr.Markdown('## 生成')
  with gr.Row():
    infer_name_input = gr.Radio(label='推理模型选择', choices=get_model_list())
    infer_txt_input = gr.Textbox(label='文本', lines=3)
    infer_audio_output = gr.Audio(type='filepath', label='为您合成的音频')
    infer_btn = gr.Button('开始语音合成', variant='primary')
    model_delete_btn = gr.Button('删除模型', variant='stop')

  # 逻辑部分
  label_btn.click(
    auto_label,
    inputs=[label_audio_input, label_name_input],
    outputs=[label_status_output, train_dataset_input]
  )

  dataset_delete_btn.click(
    delete_dataset,
    inputs=train_dataset_input,
    outputs=[train_dataset_input]
  )

  train_btn.click(
    train,
    inputs=[train_name_input, train_steps_input, train_dataset_input],
    outputs=[train_status_output, infer_name_input]
  )

  infer_btn.click(
    infer,
    inputs=[infer_name_input, infer_txt_input],
    outputs=[infer_audio_output]
  )

  model_delete_btn.click(
    delete_model,
    inputs=infer_name_input,
    outputs=[infer_name_input]
  )

  server.load(
    refresh,
    inputs=[],
    outputs=[train_dataset_input, infer_name_input]
  )


server.launch(share=True, show_error=True)