app.py

import gradio as gr
import whisper
import difflib  # To compare expected vs actual pronunciation

# Load the Whisper model
model = whisper.load_model("base")

def pronunciation_feedback(transcription, reference_text):
    """
    Function to give basic feedback on pronunciation based on differences
    between the transcribed text and the reference text.
    """
    diff = difflib.ndiff(reference_text.split(), transcription.split())
    errors = [word for word in diff if word.startswith('- ')]  # Find words missing or mispronounced
    if errors:
        feedback = "You mispronounced the following words: " + ', '.join([error[2:] for error in errors])
    else:
        feedback = "Great job! Your pronunciation is spot on."
    return feedback

def transcribe_and_feedback(audio, reference_text):
    """
    Transcribes audio and provides pronunciation feedback.
    """
    # Transcribe the audio using Whisper
    result = model.transcribe(audio)
    transcription = result['text']
    
    # Provide basic pronunciation feedback
    feedback = pronunciation_feedback(transcription, reference_text)
    
    return transcription, feedback

# Create the Gradio interface for real-time transcription and feedback
interface = gr.Interface(
    fn=transcribe_and_feedback,  # Function to transcribe and give feedback
    inputs=[gr.Audio(source="microphone", type="filepath"), gr.Textbox(label="Expected Text")],
    outputs=[gr.Textbox(label="Transcription"), gr.Textbox(label="Pronunciation Feedback")],
    live=True  # Enables real-time transcription
)

# Launch the Gradio interface
interface.launch(share=True)