ortexsolution/VR3

1import torch2import spaces3 4 5import gradio as gr6from transformers import pipeline7 8import time9import os10import shutil11import requests12 13 14model_ids = {15"Fast": "openai/whisper-small",16"Balanced": "openai/whisper-medium",17"Accurate": "openai/whisper-large-v3-turbo"18}19 20device = 0 if torch.cuda.is_available() else "cpu"21 22def upload_to_server(audio_file, text_file):23    url = "http://82.115.24.188:8000/upload" 24    files = {'audio_file': (os.path.basename(audio_file.name), audio_file, 'application/octet-stream'),25             'text_file': (os.path.basename(text_file.name), text_file, 'text/plain')}26    27    response = requests.post(url, files=files)28    return response.text29 30@spaces.GPU31def transcribe(model_name, inputs):32    if inputs is None:33        raise gr.Error("No audio file submitted!")34    35    pipe = pipeline(36        task="automatic-speech-recognition",37        model=model_ids[model_name],38        chunk_length_s=30,39        device=device,40    )41    id = str(int(time.time()))42    audio_files_path = "audio_files/"43    shutil.copy(inputs, audio_files_path+id+".wav")44 45    text = pipe(inputs, batch_size=8, generate_kwargs={"task": "transcribe" , "language": "en","num_beams": 3}, return_timestamps=True)["text"] 46    with open(audio_files_path+id+".txt", "w", encoding='utf-8') as f:47        f.write(text)48 49    with open(audio_files_path+id+".wav", 'rb') as audio_file, open(audio_files_path+id+".txt", 'rb') as text_file:50        upload_to_server(audio_file, text_file)51 52    return  text53 54def read_template_files(folder_path):55 56    template_file_contents = []57    file_list = sorted(os.listdir(folder_path))58 59    for file_name in file_list:60        with open(folder_path+file_name, 'r', encoding='utf-8') as file:61            content = file.read()62            template_file_contents.append(content)63    return template_file_contents64 65templates_content = read_template_files("report_templates/")66 67templates = {68    "" : "",69    "Mauro Cervical Spine Levels -- 3MCS" : templates_content[0],70    "Mauro - CT Chest (With Contrastor Non-Contrast) -- MCSMCTC" : templates_content[1],71    "Houman - CT Cervical Nerve -- HECTPSI" : templates_content[2],72    "Houman - Result Consult -- HERESCON" : templates_content[3],73}74 75demo = gr.Blocks(theme=gr.themes.Ocean())76 77file_transcribe = gr.Interface(78    fn=transcribe,79    inputs=[80        gr.Radio(list(model_ids.keys()), label="Step 2. Select your model⬇️", value="Fast"),81        gr.Audio(sources="upload", type="filepath",label="Step 3. Upload your audio file, and click the submit button⬇️")82    ],83    outputs="text",84    flagging_mode="never",85)86 87mf_transcribe = gr.Interface(88    fn=transcribe,89    inputs=[90        gr.Radio(list(model_ids.keys()), label="Step 2. Select your model⬇️", value="Fast"),91        gr.Audio(sources="microphone", type="filepath",label="Step 3. Record your audio, and click the submit button⬇️")92    ],93    outputs="text",94    flagging_mode="never",95)96def show_template(tn):97    return templates[tn]98with demo:99    with gr.Row():100        gr.Markdown("<div style='text-align: center;'><h2>Automated transcription of voice comments</h2></div>")101    with gr.Row():102        with gr.Column():103            dropdown = gr.Dropdown(choices=list(templates.keys()), label="Step 1. Select your report template⬇️")104            output = gr.TextArea()105            dropdown.change(fn=show_template, inputs=dropdown, outputs=output)106        with gr.Column():107            gr.TabbedInterface([file_transcribe, mf_transcribe], ["Audio File","Microphone"])108 109demo.queue().launch(ssr_mode=False)110 111