app.py

import matplotlib
matplotlib.use('TkAgg')
import gradio as gr
from time import time
import soundfile
import os
import numpy as np

names = ["助手小技，", "小技助手，", ""]

actions = ["显示", "隐藏", "全部显示", "单独显示", "仅显示", "清除", "全部清除", "打开", "关闭"]

texts1 = ['肝段', "肺段", "肝脏", "左肺", "右肺", "左三叶", "右三叶",
         "病灶", "结节", "肝肿瘤",
         "脉管", "支气管", "动脉", "静脉", "肺动脉", "肺气管", "肺静脉", "肝静脉", "门静脉", "下腔静脉", "上腔静脉", "主动脉", "食管", "胆囊管",
         "胆囊", "纵膈", "骨骼", "骨头", "心脏"]

texts2_1 = ["左肺", "右肺"]
texts2_2 = ["尖段", "后段", "前段", "外侧段", "内侧段", "背段", "内基底段", "前基底段", "外基底段", "后基底段"]

texts3_1 = ["肝脏", "右肺"]
texts3_2 = ["左外叶上段", "左外叶下段", "左内叶上段", "左内叶下段", "右前叶下段", "右后叶下段", "右后叶上段", "右前叶上段", "尾状叶"]


def update_hotapi():
    if np.random.random() > 0.6:
        times = np.random.randint(1, 6)
        new_text = np.random.choice(names) + np.random.choice(actions) + ",".join(np.random.choice(texts1, size=times, replace=False))
    else:
        if np.random.random() > 0.5:
            new_text = np.random.choice(names) + np.random.choice(actions) + np.random.choice(texts2_1) + np.random.choice(texts2_2)
        else:
            new_text = np.random.choice(names) + np.random.choice(actions) + np.random.choice(texts3_1) + np.random.choice(texts3_2)

    return new_text


cnt_dict = {}


def hot_save(audio_list, text, prefix_text):
    global cnt_dict
    if prefix_text not in cnt_dict:
        cnt_dict[prefix_text] = 0
    print("hot save start!!!!!")
    rate, audio_arr = audio_list
    wav_path = "./{}-{}-{}-cnt{}.mp3".format(prefix_text, text, str(int(time())), cnt_dict[prefix_text]))
    soundfile.write(wav_path, data=audio_arr, samplerate=rate)

    update_text = update_hotapi()

    cnt_dict[prefix_text] += 1
    print("hot save done!!!!!")

    return update_text, None, cnt_dict[prefix_text]


with gr.Blocks() as demo:
    gr.Markdown("## Smars-AI 三维重建语音**无约束**录制助手")
    gr.Markdown("#### 流程:\n"
                "- step1: 输入待采集人姓名：（随意，简称就行）\n"
                "- step2: 点击`record`/`录制`, 按照`热词采集`说话，然后点击`stop`， 因可能存在网络延迟，等待1~2s\n"
                "- step3: 点击`保存录音`，如果遇到error，再次保存；  \n")

    with gr.Group():
        text = gr.Textbox(label="热词采集:（动态随机更新）", min_width=100, value="小技助手")
        prefix_text = gr.Textbox(label="采集人,请输入：", min_width=100, value='anonymous')
        slider = gr.Slider(minimum=0, maximum=66, value=0, step=1, label="录制计数", scale=1, interactive=False)

        audio = gr.Audio(sources=["microphone"])
        save_buttom = gr.Button('保存录音')
        save_buttom.click(hot_save, inputs=[audio, text, prefix_text], outputs=[text, audio, slider])

demo.launch()