import gradio as gr
import re

# Define the regex pattern to extract text
pattern = r'text=["\'](.*?)["\'],\s*chunks=None'

# Load the Whisper model
whisper = gr.load("models/openai/whisper-large-v3-turbo")

def transcribe(audio):
    # Transcribe the audio using Whisper
    result = whisper(audio)
    
    # Use re.search to find the match
    match = re.search(pattern, result)
    
    if match:
        # Access the captured group (the content inside the quotes)
        captured_text = match.group(1)
        return captured_text
    else:
        return result

# Define the Gradio interface
gr.Interface(fn=transcribe, inputs=gr.Audio(type="filepath"), outputs=gr.Textbox()).launch()